WO2020057492A1 - Image compression and decompression method and apparatus, electronic device, and storage medium - Google Patents

Abstract

Embodiments of the present application provide an image compression and decompression method and apparatus, an electronic device, and a storage medium. The method comprises: acquiring an image to be compressed; applying a convolutional neural network model to compress said image by means of a convolution filter, and obtaining neural network-based compressed data, wherein the convolutional neural network model is trained and obtained by means of a preset training set, and the training set comprises: a plurality of low resolution images having a resolution lower than a resolution threshold; and images, corresponding to the low resolution images, having a resolution higher than the resolution threshold; and storing the neural network-based compressed data. The embodiments of the present application achieve improved image resolution for users upon viewing compressed images, and improve the user experience.

Description

Image compression and decompression method and device, electronic equipment and storage medium

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on September 19, 2018, with the application number 201811095462.0, and the invention name is "Image Compression, Decompression Method, Device, Electronic Equipment, and Storage Medium". Citations are incorporated in this application.

Technical field

The present application relates to the field of image processing technology, and in particular, to an image compression and decompression method, device, electronic device, and storage medium.

Background technique

At present, in order to reduce the storage space occupied by an image and facilitate the transmission of the image, it is often necessary to compress the size of the image. The compression of the size of the image includes reducing the size of the image to reduce the resolution of the image. For example, a 100 * 100 pixel image is compressed into a 50 * 50 pixel image. However, if the image is compressed by this compression method, when the user views the compressed image, the user can only see the compressed image, that is, the image with lower resolution, and the user experience is poor.

Summary of the Invention

The purpose of the embodiments of the present application is to provide an image compression and decompression method, device, electronic device, and storage medium, so as to improve the resolution of an image viewed by a user when viewing the compressed image, and improve the user experience. Specific technical solutions are as follows:

In a first aspect, an embodiment of the present application provides an image compression method, where the method includes:

Obtain the image to be compressed;

Use a first CNN (Convolutional Neural Network) model to perform convolution filtering and compression on the to-be-compressed image to obtain network compressed data; wherein the first CNN model and the second CNN model use preset training The training set is obtained. The training set includes multiple low-resolution images with a resolution lower than a resolution threshold, and images with a resolution higher than the resolution threshold corresponding to the low-resolution images. The second CNN model is used for Performing convolution filtering decompression on the network compressed data to obtain a decompressed image;

The network compressed data is stored.

Optionally, the method further includes:

After obtaining the image to be compressed, performing size compression on the image to be compressed to obtain a size-compressed image;

After the network compressed data is obtained, the correspondence between the size compressed image and a preset identifier is stored; the preset identifier is used to indicate the presence of the network compressed data of the image to be compressed.

Optionally, the image to be compressed is located in a PDF document; or the image to be compressed is located in a Word document; or the image to be compressed is located in an Excel document; or the image to be compressed is located in a WPS document.

In a second aspect, an embodiment of the present application provides an image decompression method, where the method includes:

Obtain the network compressed data of the target image;

Convolutional filtering and decompression of the network compressed data using a second CNN model to obtain a decompressed image; wherein the second CNN model and the first CNN model are obtained by training using a preset training set, and the training set includes A plurality of low-resolution images with a resolution lower than a resolution threshold, and images corresponding to a low-resolution image with a resolution higher than the resolution threshold, the first CNN model is used to perform convolution filtering on image compression Compress to obtain the network compressed data;

The decompressed image is output.

Optionally, the step of obtaining network compressed data of a target image includes:

Obtain the size-compressed image of the target image;

Determining whether a preset identifier corresponding to the compressed image of the size is stored; the preset identifier is used to indicate that network compression data of the target image exists;

If stored, the network compressed data is obtained.

Optionally, the method further includes:

If the preset identifier corresponding to the size-compressed image is not stored, the size-compressed image is output.

In a third aspect, an embodiment of the present application provides an image compression apparatus, where the apparatus includes:

An acquisition module for acquiring an image to be compressed;

A first compression module, configured to perform convolution filtering and compression on the image to be compressed using a first CNN model to obtain network compressed data; wherein the first CNN model and the second CNN model are obtained by training using a preset training set The training set includes multiple low-resolution images with a resolution lower than the resolution threshold, and images with a resolution higher than the resolution threshold corresponding to the low-resolution images, and the second CNN model is used for Decompress the network compressed data by convolution filtering to obtain a decompressed image;

The first storage module is configured to store the network compressed data.

Optionally, the device further includes:

A second compression module, configured to compress the image to be compressed after obtaining the image to be compressed to obtain a size-compressed image;

A second storage module is configured to store the correspondence between the size compressed image and a preset identifier after the network compressed data is obtained, where the preset identifier is used to indicate the existence of network compressed data of the image to be compressed.

Optionally, the image to be compressed is located in a PDF document; or the image to be compressed is located in a Word document; or the image to be compressed is located in an Excel document; or the image to be compressed is located in a WPS document.

In a fourth aspect, an embodiment of the present application provides an image decompression device, where the device includes:

An acquisition module for acquiring network compressed data of a target image;

A decompression module, configured to perform convolution filtering and decompression on the network compressed data using a second CNN model to obtain a decompressed image; wherein the second CNN model and the first CNN model are obtained by training using a preset training set; The training set includes multiple low-resolution images with a resolution lower than a resolution threshold, and images with a resolution higher than the resolution threshold corresponding to the low-resolution images, and the first CNN model is used to be compressed. Image compression and convolution filter compression to obtain the network compression data;

An output module, configured to output the decompressed image.

Optionally, the obtaining module is specifically configured to obtain a size-compressed image of the target image; determine whether a preset identifier corresponding to the size-compressed image is stored; and the preset identifier is used to indicate a network in which the target image exists Compressed data; if stored, obtain the network compressed data.

Optionally, the output module is further configured to output the size-compressed image if the preset identifier corresponding to the size-compressed image is not stored.

In a fifth aspect, an embodiment of the present application provides an image compression method, where the method includes:

Obtain the image to be compressed;

The first CNN model is used to perform convolution filtering and compression on the image to be compressed to obtain network compressed data; wherein the first CNN model and the second CNN model are obtained by training using a preset training set, and the training set includes multiple Sample images, the second CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image;

The network compressed data is stored.

According to a sixth aspect, an embodiment of the present application provides an image decompression method, where the method includes:

Obtain the network compressed data of the target image;

Convolutional filtering and decompression of the network compressed data using a second CNN model to obtain a decompressed image; wherein the second CNN model and the first CNN model are obtained by training using a preset training set, and the training set includes Multiple sample images, the first CNN model is used to perform convolution filter compression on the image to be compressed to obtain the network compression data;

The decompressed image is output.

In a seventh aspect, an embodiment of the present application provides an image compression apparatus, where the apparatus includes:

An acquisition module for acquiring an image to be compressed;

A first compression module, configured to perform convolution filtering and compression on the image to be compressed using a first CNN model to obtain network compressed data; wherein the first CNN model and the second CNN model are obtained by training using a preset training set The training set includes multiple sample images, and the second CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image;

The first storage module is configured to store the network compressed data.

In an eighth aspect, an embodiment of the present application provides an image decompression device, where the device includes:

An acquisition module for acquiring network compressed data of a target image;

A decompression module, configured to perform convolution filtering and decompression on the network compressed data using a second CNN model to obtain a decompressed image; wherein the second CNN model and the first CNN model are obtained by training using a preset training set; The training set includes multiple sample images, and the first CNN model is used to perform convolution filter compression on the compressed image to be compressed to obtain the network compressed data;

An output module, configured to output the decompressed image.

In a ninth aspect, an embodiment of the present application provides an electronic device including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete each other through the communication bus. Communication;

The memory is used to store a computer program;

The processor is configured to execute a program stored on the memory to implement any of the steps of the image compression method provided by the first aspect.

According to a tenth aspect, an embodiment of the present application provides an electronic device including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete each other through the communication bus. Communication;

The memory is used to store a computer program;

The processor is configured to execute a program stored on the memory to implement any of the steps of the image decompression method provided by the second aspect.

According to an eleventh aspect, an embodiment of the present application provides an electronic device including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete each other through the communication bus Communication

The memory is used to store a computer program;

The processor is configured to execute a program stored on the memory to implement any of the steps of the image compression method provided by the fifth aspect.

In a twelfth aspect, an embodiment of the present application provides an electronic device including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete each other through the communication bus. Communication

The memory is used to store a computer program;

The processor is configured to execute a program stored on the memory to implement any of the steps of the image decompression method provided by the sixth aspect.

In a thirteenth aspect, an embodiment of the present application provides a machine-readable storage medium. The machine-readable storage medium stores a computer program, and when the computer program is executed by a processor, any image provided in the first aspect is implemented. Compression method steps.

In a fourteenth aspect, an embodiment of the present application provides a machine-readable storage medium. The machine-readable storage medium stores a computer program. When the computer program is executed by a processor, any image provided in the second aspect is implemented. Decompression method steps.

In a fifteenth aspect, an embodiment of the present application provides a machine-readable storage medium. The machine-readable storage medium stores a computer program, and when the computer program is executed by a processor, any image provided in the fifth aspect is implemented. Compression method steps.

In a sixteenth aspect, an embodiment of the present application provides a machine-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, any image provided in the sixth aspect is implemented Decompression method steps.

In a seventeenth aspect, an embodiment of the present application provides a computer program that, when executed by a processor, implements any of the image compression method steps provided in the first aspect.

In an eighteenth aspect, an embodiment of the present application provides a computer program that, when executed by a processor, implements any of the image decompression method steps provided in the second aspect.

In a nineteenth aspect, an embodiment of the present application provides a computer program that, when executed by a processor, implements any of the image compression method steps provided in the fifth aspect.

In a twentieth aspect, an embodiment of the present application provides a computer program that, when executed by a processor, implements any of the image decompression method steps provided in the sixth aspect.

Embodiments of the present application provide an image compression and decompression method, device, electronic device, and storage medium. A CNN model is trained by using multiple images below a resolution threshold and multiple images above a resolution threshold. The CNN model is used to perform convolution filtering and compression on the images to obtain network compression data with smaller resolution to reduce the storage space occupied. Then, the CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image. Since the CNN model is trained using multiple images below the resolution threshold and multiple images above the resolution threshold, the CNN model can be used to obtain a higher resolution decompressed image, and even the resolution of the decompressed image The rate is higher than the resolution of the original image. This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience. Of course, to implement any product or method of the present application, it is not necessary to achieve all the advantages described above at the same time.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to explain the technical solutions in the embodiments of the present application or the prior art more clearly, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely These are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained according to these drawings without paying creative labor.

FIG. 1 is a first schematic flowchart of an image compression method according to an embodiment of the present application; FIG.

2 is a schematic flowchart of a CNN model training method according to an embodiment of the present application;

3 is a schematic flowchart of a second method of an image compression method according to an embodiment of the present application;

4 is a first schematic flowchart of an image decompression method according to an embodiment of the present application;

FIG. 5 is a second schematic flowchart of an image decompression method according to an embodiment of the present application; FIG.

FIG. 6 is a first schematic structural diagram of an image compression device according to an embodiment of the present application; FIG.

7 is a schematic diagram of a second structure of an image compression device according to an embodiment of the present application;

8 is a first schematic structural diagram of an image decompression device according to an embodiment of the present application;

FIG. 9 is a first schematic structural diagram of an electronic device according to an embodiment of the present application; FIG.

FIG. 10 is a schematic diagram of a second structure of an electronic device according to an embodiment of the present application.

detailed description

In the following, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without doing creative work fall within the protection scope of the present application.

In order to reduce the loss of image resolution when viewing the compressed image and improve the user experience, embodiments of the present application provide an image compression method and an image decompression method. The image compression and decompression method can be applied to any electronic device such as a mobile phone, a computer, and a notebook.

In the image compression and decompression method provided by the embodiments of the present application, a CNN model is trained by using multiple images below a resolution threshold and multiple images above a resolution threshold. The CNN model is used to perform convolution filtering and compression on the image to obtain network compression data with a smaller resolution to reduce the occupied storage space. Then, the CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image. Since the CNN model is trained using multiple images below the resolution threshold and multiple images above the resolution threshold, the CNN model can be used to obtain a higher resolution decompressed image, and even the resolution of the decompressed image The rate is higher than the resolution of the original image. This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

The image compression method and the image decompression method provided by the embodiments of the present application will be described in detail below through specific embodiments.

Referring to FIG. 1, FIG. 1 is a schematic flowchart of a first method of an image compression method according to an embodiment of the present application. The method includes the following steps.

Step 101: Obtain an image to be compressed.

When an image needs to be compressed, the image is acquired as an image to be compressed.

In the embodiment of the present application, the image to be compressed may be an independent image or an image located in a document. The documents here include but are not limited to PDF documents, Word documents, Excel documents, WPS documents, etc. The PDF document may be an editable PDF document.

Step 102: Compress the compressed image using the CNN model to obtain network compressed data.

Step 102 is to use the first CNN model to perform convolution filtering and compression on the image to be compressed to obtain network compressed data.

The first CNN model is used to perform convolution filter compression on the image to be compressed to obtain network compressed data. The first CNN model and the second CNN model are obtained by training with a preset training set. The training set includes multiple low-resolution images with a resolution lower than the resolution threshold, and the resolution corresponding to the low-resolution image is higher than the resolution threshold Image, the second CNN model is used to decompress the network compressed data by convolution filtering to obtain a decompressed image.

In the embodiment of the present application, an image with a resolution higher than the resolution threshold is a high-resolution image. A low-resolution image may correspond to one or more high-resolution images, and a high-resolution image may correspond to one or more low-resolution images.

In an embodiment of the present application, referring to the training process of the first CNN model shown in FIG. 2, the method includes the following steps.

Step 21: Obtain a preset first CNN model. Initialize the parameters in the first CNN model. The initialized parameters can be set according to actual needs and experience.

In this step, training-related high-level parameters may also be set, where the high-level parameters may include a learning rate, a gradient descent algorithm, and the like. Specifically, the high-level parameters may be set in various manners in related technologies, which is not described in detail here.

Step 22: Obtain a preset training set. The training set includes multiple low-resolution images with a resolution below the resolution threshold, and high-resolution images corresponding to the low-resolution image and with a resolution above the resolution threshold.

Step 23: Obtain network compressed data corresponding to multiple high-resolution images.

The network compressed data of the high-resolution image can be stored in a storage database in advance. The network compression data of the high-resolution image is input to the second CNN model, and the high-resolution image can be obtained.

Step 24: Perform forward calculation, which specifically includes: inputting each low-resolution image in the training set into a preset first CNN model and performing convolution filtering and compression processing to obtain network compressed data corresponding to each low-resolution image.

The execution order of steps 23 and 24 is not limited in the embodiment of the present application.

Step 25: Determine a loss value of image convolution filter compression based on network compressed data corresponding to multiple high-resolution images and network compressed data corresponding to each low-resolution image.

In the embodiment of the present application, the similarity between the network compressed data corresponding to the high-resolution image and the network compressed data corresponding to the low-resolution image can be calculated, and the reciprocal of the calculated similarity is used as the loss value of the image convolution filter compression. The larger the similarity, the smaller the loss value of image convolution filter compression. The smaller the similarity, the larger the loss value of image convolution filter compression.

In the embodiment of the present application, a Mean Squared Error (MSE) formula can also be used as a loss function to calculate the mean square error of the network compressed data corresponding to the high-resolution image and the network compressed data corresponding to the low-resolution image. The loss value of image convolution filter compression is obtained. For details, refer to the description of MSE in related technologies, and details are not described herein again.

Step 26: Determine whether the preset first CNN model converges based on the loss value of the image convolution filter compression. If not, proceed to step 27; if converge, end the first CNN model training.

In one embodiment, a first loss threshold may be set in advance. If the loss value of the image convolution filter compression is lower than the first loss threshold, it is determined that the first CNN model converges. If the loss value of the image convolution filter compression is not lower than the first loss threshold, it is determined that the first CNN model does not converge.

Step 27: Adjust parameters in the preset first CNN model, and return to step 24.

The electronic device for training the first CNN model and the electronic device for compressing images may be the same device or different devices.

Step 103: Store network compression data.

In the embodiment of the present application, if the image to be compressed is an independent image, after obtaining the network compressed data, the obtained network compressed data is directly stored. If the image to be compressed is an image located in a document, after obtaining the network compressed data, the obtained network compressed data is stored in the document. For example, if the image 1 to be compressed is located in the PDF document f1, after obtaining the network compressed data 1 of the image 1 to be compressed, the network compressed data 1 is stored in the PDF document f1.

In the embodiment of the present application, the network compressed data obtained by using the first CNN model compression is not an image, and image processing software cannot directly open the network compressed data to obtain an image. For ease of reference, reference is made to FIG. 3, which is a schematic flowchart of a second method for providing an image compression method according to an embodiment of the present application. Based on FIG. 1, the method includes the following steps.

Step 301: Obtain an image to be compressed.

Step 301 is the same as step 101.

Step 302: Compress the image to be compressed using the CNN model to obtain network compressed data.

Step 302 is to use the first CNN model to perform convolution filtering and compression on the image to be compressed to obtain network compressed data.

The first CNN model is used to perform convolution filter compression on the image to be compressed to obtain network compressed data. The first CNN model and the second CNN model are obtained by training with a preset training set. The training set includes multiple low-resolution images with a resolution lower than the resolution threshold, and the resolution corresponding to the low-resolution image is higher than the resolution threshold. Image, the second CNN model is used to decompress the network compressed data by convolution filtering to obtain a decompressed image.

Step 302 is the same as step 102.

Step 303: Size-compress the image to be compressed to obtain a size-compressed image.

For example, size compression is performed on an image of 100 * 100 pixels to obtain an image of 50 * 50 pixels. The 50 * 50 pixel image is a size-compressed image.

Step 304: Store the network compressed data, and store the correspondence between the size compressed image and the preset identifier. The preset identifier is used to indicate that there is network compressed data of the image to be compressed.

In the embodiment of the present application, after obtaining the size-compressed image, the size-compressed image may be stored first, and after obtaining the network-compressed data, the preset identifier corresponding to the size-compressed image may be stored. It is also possible to first obtain network compressed data, and after obtaining the size compressed image, directly store the correspondence between the size compressed image and the preset identifier. This embodiment of the present application does not limit this.

In the embodiment of the present application, both network compressed data and size compressed images are stored. If the second CNN model is not stored in the electronic device, the image cannot be opened through the network compressed data. At this time, the size compressed image can be opened to Avoid issues where users cannot view images.

Based on the foregoing embodiment of the image compression method, an embodiment of the present application further provides an image decompression method. Referring to FIG. 4, FIG. 4 is a first schematic flowchart of an image decompression method provided by an embodiment of the present application. The method includes the following steps.

Step 401: Obtain network compressed data of a target image.

In the embodiment of the present application, the target image is an image that the user needs to open. The target image can be an independent image or an image located in a document. The documents here include but are not limited to PDF documents, Word documents, Excel documents, WPS documents.

If the user needs to open a document, all the images in the document can be used as the target image.

Step 402: Decompress the network compressed data using a CNN model to obtain a decompressed image.

Step 402 is to use the second convolutional neural network model to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image.

The second CNN model is used to perform convolution filtering and decompression on network compressed data to obtain a decompressed image. The second CNN model and the first CNN model are obtained by training using a preset training set. The training set includes multiple low-resolution images with a resolution lower than the resolution threshold, and the resolution corresponding to the low-resolution image is higher than the resolution threshold. Image, the first CNN model is used to perform convolution filter compression on the image to be compressed to obtain network compressed data.

In an embodiment of the present application, the training of the second CNN model may refer to the training process of the first CNN model shown in FIG. 2. The difference between the training process of the second CNN model and the training process of the first CNN model lies in: obtaining network compression data corresponding to multiple low-resolution images, and inputting network compression data corresponding to each low-resolution image in the training set into the The two CNN models perform convolution filtering and decompression processing to obtain decompressed images. Based on multiple high-resolution images and decompressed images, the loss value of image convolution filtering and decompression is determined.

The electronic device for training the second CNN model and the electronic device for decompressing the image may be the same device or different devices.

Step 403: Output a decompressed image.

In the embodiment of the present application, the second CNN model is trained by using multiple images below the resolution threshold and multiple images above the resolution threshold, so that the second CNN model is more suitable for decompressing network compressed data into an image. Decompressed images with higher resolution are obtained, and even decompressed images with higher resolution than the original image (such as low-resolution images) are obtained. This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

In the embodiment of the present application, the network compressed data is data obtained by using convolution filter compression. Therefore, the network compressed data is not image data. Using image processing software cannot directly open the network compressed data to obtain an image. For ease of reference, reference is made to FIG. 5, which is a schematic flowchart of a second method of the image decompression method provided by an embodiment of the present application. Based on FIG. 4, the method includes the following steps.

Step 501: Obtain a size-compressed image of a target image.

Step 502: Determine whether a preset identifier corresponding to the size-compressed image is stored. If it is stored, step 503 is performed. If not, step 506 is performed. The preset identifier is used to indicate that there is network compressed data of the target image.

In the embodiment of the present application, if a preset identifier corresponding to the size compressed image is stored, it is determined that network compression data of the target image exists. In order to improve the resolution of the image viewed by the user when viewing the compressed image, step 503 is performed. If the preset identifier corresponding to the size compressed image is not stored, it is determined that there is no network compressed data of the target image, and step 506 is performed to ensure that the user can view the image.

Step 503: Obtain network compressed data.

Step 504: Decompress the network compressed data by using a CNN model to obtain a decompressed image.

Step 504 is to use the second convolutional neural network model to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image.

The second CNN model is used to perform convolution filtering and decompression on network compressed data to obtain a decompressed image. The second CNN model and the first CNN model are obtained by training using a preset training set. The training set includes multiple low-resolution images with a resolution lower than the resolution threshold, and the resolution corresponding to the low-resolution image is higher than the resolution threshold Image, the first CNN model is used to perform convolution filter compression on the image to be compressed to obtain network compressed data.

Step 505: Output a decompressed image.

Steps 503-505 are the same as steps 401-403.

Step 506: Output the compressed image.

In an embodiment of the present application, if a decompressed image cannot be obtained, for example, because there is no second CNN model, network compression data cannot be decompressed to obtain a decompressed image, a size-compressed image is directly output to ensure the user View the image.

In an embodiment of the present application, the first CNN model and the second CNN model may be jointly trained. Specifically, the joint training of the first CNN model and the second CNN model may include the following steps.

Step a1: Obtain a preset first CNN model and a second CNN model. Initialize the parameters in the first CNN model and the second CNN model. The initialized parameters can be set according to actual needs and experience.

Step a2: Obtain a preset training set. The training set includes multiple low-resolution images with a resolution below the resolution threshold, and high-resolution images corresponding to the low-resolution image and with a resolution above the resolution threshold.

Step a3. Perform the first forward calculation, which specifically includes: inputting each low-resolution image in the training set into a preset first CNN model and performing convolution filtering and compression processing to obtain a network corresponding to each low-resolution image. Compressed data.

In an optional embodiment, after obtaining network compressed data corresponding to a low-resolution image, it is determined that the data amount of the network compressed data is less than a preset data amount. If it is less than the preset data amount, step a4 is performed. If it is greater than or equal to the preset data amount, step a7 is performed.

Step a4, performing a second forward calculation, which specifically includes: inputting network compression data corresponding to each low-resolution image into a preset second CNN model and performing convolution filtering and decompression processing to obtain each low-resolution image The corresponding decompressed image.

Step a5: Determine an image loss value based on the high-resolution image and the decompressed image corresponding to each low-resolution image.

In the embodiment of the present application, the similarity between the high-resolution image and the decompressed image corresponding to the low-resolution image may be calculated, and the reciprocal of the calculated similarity is used as the image loss value. The greater the similarity, the smaller the image loss value. The smaller the similarity, the larger the image loss value.

In the embodiment of the present application, the MSE formula can also be used as a loss function to calculate the mean square error between the high-resolution image and the decompressed image corresponding to the low-resolution image to obtain the image loss value. For details, refer to the description of MSE in related technologies, and details are not described herein again.

Step a6: Determine whether the preset first CNN model and the second CNN model converge based on the image loss value. If not, proceed to step a7; if converge, end training.

In one embodiment, a second loss threshold may be set in advance. If the image loss value is lower than the second loss threshold, it is determined that the first CNN model and the second CNN model converge. If the image loss value is not lower than the second loss threshold, it is determined that the first CNN model and the second CNN model do not converge.

Step a7: Adjust the parameters in the preset first CNN model and the second CNN model, and return to step a3.

The compressed image of the first CNN model obtained by training in steps a1 to a7 above is used to obtain network compressed data. The amount of data of the network compressed data is less than that of the image to be compressed, which reduces the storage space occupied by the image. In addition, the second CNN model trained in the above steps a1-a7 is used to decompress the network compressed data to obtain a decompressed image. Since the second CNN model obtained by training the low-resolution image and the high-resolution image is used, the resolution of the decompressed image is higher than or equal to that of the image to be compressed (that is, the original image). This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

Based on the foregoing embodiment of the image compression method, an embodiment of the present application further provides an image compression device. Referring to FIG. 6, FIG. 6 is a schematic diagram of a first structure of an image compression apparatus according to an embodiment of the present application. The apparatus includes the following modules.

An acquisition module 601, configured to acquire an image to be compressed;

A first compression module 602 is configured to perform convolution filtering and compression on an image to be compressed by using a first CNN model to obtain network compressed data. The first CNN model and the second CNN model are obtained by training using a preset training set. Multiple low-resolution images with a resolution lower than the resolution threshold and images with a resolution higher than the resolution threshold corresponding to the low-resolution image. The second CNN model is used to perform convolution filtering and decompression on network compressed data to obtain Decompress the image;

The first storage module 603 is configured to store network compressed data.

In an embodiment of the present application, referring to the image compression device shown in FIG. 7, based on FIG. 6, it may further include:

A second compression module 604, configured to perform size compression on the image to be compressed after obtaining the image to be compressed to obtain a size-compressed image;

The second storage module 605 is configured to store the correspondence between the size-compressed image and a preset identifier after the network compressed data is obtained; the preset identifier is used to indicate that there is network compressed data of the image to be compressed.

In an embodiment of the present application, the image to be compressed is located in a PDF document; or

The image to be compressed is in a Word document; or

The image to be compressed is in an Excel document; or

The image to be compressed is located in a WPS document.

An embodiment of the present application provides an image compression apparatus, which uses a plurality of images below a resolution threshold and a plurality of images above a resolution threshold to train a CNN model. The CNN model is used to perform convolution filtering and compression on the image to obtain network compression data with a smaller resolution to reduce the occupied storage space. Then, the CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image. Since the CNN model is trained using multiple images below the resolution threshold and multiple images above the resolution threshold, the CNN model can be used to obtain a higher resolution decompressed image, and even the resolution of the decompressed image The rate is higher than the resolution of the original image. This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

Based on the foregoing embodiment of the image decompression method, an embodiment of the present application further provides an image decompression device. Referring to FIG. 8, FIG. 8 is a schematic diagram of a first structure of an image decompression device provided by an embodiment of the present application. The device includes the following modules.

An obtaining module 801, configured to obtain network compressed data of a target image;

The decompression module 802 is configured to perform convolution filtering and decompression on the network compressed data by using the second CNN model to obtain a decompressed image. The second CNN model and the first CNN model are obtained by training using a preset training set. Including multiple low-resolution images with a resolution lower than the resolution threshold and images with a resolution higher than the resolution threshold corresponding to the low-resolution image, the first CNN model is used to perform convolution filter compression on the image to be compressed to obtain Network compressed data;

An output module 803 is configured to output a decompressed image.

In an embodiment of the present application, the obtaining module 801 may be specifically configured to obtain a size-compressed image of a target image; determine whether a preset identifier corresponding to the size-compressed image is stored; the preset identifier is used to indicate that network compression of the target image exists Data; if stored, get network compressed data.

In an embodiment of the present application, the output module 803 may be further configured to output a size-compressed image if a preset identifier corresponding to the size-compressed image is not stored.

An embodiment of the present application provides an image decompression apparatus, which uses a plurality of images below a resolution threshold and a plurality of images above a resolution threshold to train a CNN model. The CNN model is used to perform convolution filtering and compression on the image to obtain network compression data with a smaller resolution to reduce the occupied storage space. Then, the CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image. Since the CNN model is trained using multiple images below the resolution threshold and multiple images above the resolution threshold, the CNN model can be used to obtain a higher resolution decompressed image, and even the resolution of the decompressed image The rate is higher than the resolution of the original image. This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

An embodiment of the present application further provides an image compression method. The method includes the following steps.

Step 110: Obtain an image to be compressed.

Step 120: Use the first CNN model to perform convolution filtering and compression on the compressed image to obtain network compression data. The first CNN model and the second CNN model are obtained by training using a preset training set, where the training set includes multiple sample images The second CNN model is used to perform convolution filtering and decompression on network compressed data to obtain a decompressed image.

Step 130: Store network compression data.

In an embodiment of the present application, the image compression method may further include:

After obtaining the image to be compressed, the image to be compressed is subjected to size compression to obtain a size-compressed image;

After the network compressed data is obtained, the correspondence between the size-compressed image and the preset identifier is stored; the preset identifier is used to indicate that there is network compressed data of the image to be compressed.

In one embodiment of the present application, the image to be compressed is located in a PDF document; or the image to be compressed is located in a Word document; or the image to be compressed is located in an Excel document; or the image to be compressed is located in a WPS document.

An embodiment of the present application provides an image compression method. A CNN model is trained by using multiple images below a resolution threshold and multiple images above a resolution threshold. The CNN model is used to perform convolution filtering and compression on the image to obtain network compression data with a smaller resolution to reduce the occupied storage space. Then, the CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image. Since the CNN model is trained using multiple images below the resolution threshold and multiple images above the resolution threshold, the CNN model can be used to obtain a higher resolution decompressed image, and even the resolution of the decompressed image The rate is higher than the resolution of the original image. This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

An embodiment of the present application further provides an image decompression method. The method includes the following steps.

Step 210: Obtain network compressed data of the target image.

Step 220: Use the second CNN model to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; wherein the second CNN model and the first CNN model are obtained by training using a preset training set, and the training set includes multiple samples Image, the first CNN model is used to perform convolution filter compression on the image to be compressed, to obtain network compressed data.

Step 230: Output a decompressed image.

In an embodiment of the present application, the step of obtaining network compressed data of a target image may include:

Obtain the size-compressed image of the target image;

Determine whether a preset identifier corresponding to the size compressed image is stored; the preset identifier is used to indicate the presence of network compressed data of the target image;

If it is stored, the network compressed data is obtained.

In an embodiment of the present application, the image decompression method may further include:

If the preset identifier corresponding to the size-compressed image is not stored, the size-compressed image is output.

An embodiment of the present application provides an image decompression method, which uses multiple images below a resolution threshold and multiple images above a resolution threshold to train a CNN model. The CNN model is used to perform convolution filtering and compression on the image to obtain network compression data with a smaller resolution to reduce the occupied storage space. Then, the CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image. Since the CNN model is trained using multiple images below the resolution threshold and multiple images above the resolution threshold, the CNN model can be used to obtain a higher resolution decompressed image, and even the resolution of the decompressed image The rate is higher than the resolution of the original image. This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

In an embodiment of the present application, the first CNN model and the second CNN model may be jointly trained. Specifically, the joint training of the first CNN model and the second CNN model may include the following steps.

Step b1: Obtain a preset first CNN model and a second CNN model. Initialize the parameters in the first CNN model and the second CNN model. The initialized parameters can be set according to actual needs and experience.

Step b2: Obtain a preset training set. The training set includes multiple sample images.

Step b3: Perform the first forward calculation, which specifically includes: inputting each sample image in the training set into a preset first CNN model and performing convolution filtering and compression processing to obtain network compression data corresponding to each sample image.

In an optional embodiment, after obtaining network compressed data corresponding to a sample image, it is determined that the data amount of the network compressed data is less than a preset data amount. If it is less than the preset data amount, step b4 is performed. If it is greater than or equal to the preset data amount, step b7 is performed.

Step b4, performing a second forward calculation, which specifically includes: inputting network compression data corresponding to each sample image into a preset second CNN model and performing convolution filtering and decompression processing to obtain a decompression corresponding to each sample image. image.

Step b5: Determine an image loss value based on each sample image and the corresponding decompressed image.

In the embodiment of the present application, the similarity between the sample image and the corresponding decompressed image may be calculated, and the reciprocal of the calculated similarity is used as the image loss value. The greater the similarity, the smaller the image loss value. The smaller the similarity, the larger the image loss value.

In the embodiment of the present application, the MSE formula can also be used as a loss function to calculate the mean square error between the sample image and the corresponding decompressed image to obtain the image loss value. For details, refer to the description of MSE in related technologies, and details are not described herein again.

Step b6: Determine whether the preset first CNN model and the second CNN model converge based on the image loss value. If not, proceed to step b7; if converge, end training.

In one embodiment, a third loss threshold may be set in advance. If the image loss value is lower than the third loss threshold, it is determined that the first CNN model and the second CNN model converge. If the image loss value is not lower than the third loss threshold, it is determined that the first CNN model and the second CNN model do not converge.

Step b7: Adjust the parameters in the preset first CNN model and the second CNN model, and return to step b3.

The first CNN model obtained by training in steps b1 to b7 is used to compress the image to obtain network compressed data. The amount of network compressed data is less than the amount of data to be compressed, which reduces the storage space occupied by the image. In addition, the second CNN model trained in the above steps b1-b7 is used to decompress the network compressed data to obtain a decompressed image. Because the second CNN model is obtained by training with multiple sample images, the resolution of the decompressed image is relatively close to the resolution of the image to be compressed (that is, the original image). This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

An embodiment of the present application further provides an image compression device. The device includes an acquisition module 310, a first compression module 320, and a first storage module 330.

An obtaining module 310, configured to obtain an image to be compressed;

A first compression module 320 is configured to perform convolution filtering and compression on a compressed image using a first CNN model to obtain network compressed data; wherein the first CNN model and the second CNN model are obtained by training using a preset training set, and the training set includes Multiple sample images, the second CNN model is used to decompress the network compressed data by convolution filtering to obtain the decompressed image;

The first storage module 330 is configured to store network compressed data.

In an embodiment of the present application, the image compression apparatus may further include:

A second compression module, configured to compress the size of the image to be compressed after obtaining the image to be compressed to obtain a size-compressed image;

The second storage module is configured to store the correspondence between the size-compressed image and the preset identifier after the network compressed data is obtained; the preset identifier is used to indicate that there is network compressed data of the image to be compressed.

In one embodiment of the present application, the image to be compressed is located in a PDF document; or the image to be compressed is located in a Word document; or the image to be compressed is located in an Excel document; or the image to be compressed is located in a WPS document.

An embodiment of the present application provides an image compression device, which trains a CNN model by using multiple images below a resolution threshold and multiple images above a resolution threshold. The CNN model is used to perform convolution filtering and compression on the image to obtain network compression data with a smaller resolution to reduce the occupied storage space. Then, the CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image. Since the CNN model is trained using multiple images below the resolution threshold and multiple images above the resolution threshold, the CNN model can be used to obtain a higher resolution decompressed image, and even the resolution of the decompressed image The rate is higher than the resolution of the original image. This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

An embodiment of the present application further provides an image decompression device. The device includes: an acquisition module 410, a decompression module 420, and an output module 430.

An obtaining module 410, configured to obtain network compressed data of a target image;

The decompression module 420 is configured to perform convolution filtering and decompression on the network compressed data by using the second CNN model to obtain a decompressed image. The second CNN model and the first CNN model are obtained by training using a preset training set. Including multiple sample images, the first CNN model is used to perform convolution filter compression on the compressed image to be compressed to obtain network compressed data;

The output module 430 is configured to output a decompressed image.

In an embodiment of the present application, the obtaining module 410 may be specifically configured to obtain the size-compressed image of the target image; determine whether a preset identifier corresponding to the size-compressed image is stored; the preset identifier is used to indicate that network compression of the target image exists Data; if stored, get network compressed data.

In an embodiment of the present application, the output module 430 may be further configured to output a size-compressed image if a preset identifier corresponding to the size-compressed image is not stored.

An embodiment of the present application provides an image decompression device, which uses a plurality of images below a resolution threshold and a plurality of images above a resolution threshold to train a CNN model. The CNN model is used to perform convolution filtering and compression on the image to obtain network compression data with a smaller resolution to reduce the occupied storage space. Then, the CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image. Since the CNN model is trained using multiple images below the resolution threshold and multiple images above the resolution threshold, the CNN model can be used to obtain a higher resolution decompressed image, and even the resolution of the decompressed image The rate is higher than the resolution of the original image. This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

Based on the foregoing embodiment of the image compression method, an embodiment of the present application further provides an electronic device, as shown in FIG. 9, including an processor 901, a communication interface 902, a memory 903, and a communication bus 904. Among them, the processor 901, the communication interface 902. The memory 903 completes communication with each other through the communication bus 904.

The memory 903 is configured to store a computer program;

The processor 901 is configured to implement the foregoing image compression method when executing a program stored in the memory 903. Among them, the image compression method includes:

Obtain the image to be compressed;

The first CNN model is used to perform convolution filtering and compression on the compressed image to obtain network compressed data. The first CNN model and the second CNN model are obtained by training using a preset training set, and the training set includes multiple resolutions lower than the resolution. Threshold low-resolution images, and images corresponding to low-resolution images with a resolution higher than the resolution threshold, the second CNN model is used to perform convolution filtering and decompression on network compressed data to obtain decompressed images;

Storage network compresses data.

An embodiment of the present application provides an electronic device for training a CNN model by using multiple images below a resolution threshold and multiple images above a resolution threshold. The CNN model is used to perform convolution filtering and compression on the image to obtain network compression data with a smaller resolution to reduce the occupied storage space. Then, the CNN model is used to decompress the network compressed data into a convolution filter to obtain a decompressed image. Since the CNN model is trained using multiple images below the resolution threshold and multiple images above the resolution threshold, the CNN model can be used to obtain a higher resolution decompressed image, and even the resolution of the decompressed image The rate is higher than the resolution of the original image. This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

Based on the foregoing embodiment of the image decompression method, an embodiment of the present application further provides an electronic device. As shown in FIG. 10, the electronic device includes a processor 1001, a communication interface 1002, a memory 1003, and a communication bus 1004. The processor 1001, The interface 1002 and the memory 1003 complete communication with each other through the communication bus 1004.

A memory 1003, configured to store a computer program;

The processor 1001 is configured to implement the foregoing image decompression method when executing a program stored in the memory 1003. The image decompression method includes:

Obtain the network compressed data of the target image;

The second CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; wherein the second CNN model and the first CNN model are obtained by training using a preset training set, and the training set includes multiple resolutions lower than The low-resolution image with a resolution threshold and the image with a resolution higher than the resolution threshold corresponding to the low-resolution image, the first CNN model is used to perform convolution filter compression on the compressed image compression to obtain network compressed data;

Output decompressed image.

An embodiment of the present application provides an electronic device for training a CNN model by using multiple images below a resolution threshold and multiple images above a resolution threshold. The CNN model is used to perform convolution filtering and compression on the image to obtain network compression data with a smaller resolution to reduce the occupied storage space. Then, the CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image. Since the CNN model is trained using multiple images below the resolution threshold and multiple images above the resolution threshold, the CNN model can be used to obtain a higher resolution decompressed image, and even the resolution of the decompressed image The rate is higher than the resolution of the original image. This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

Based on the foregoing embodiment of the image compression method, an embodiment of the present application further provides an electronic device including a processor, a communication interface, a memory, and a communication bus. The processor, the communication interface, and the memory always perform mutual communication through communication.

Memory for storing computer programs;

The processor is configured to implement the foregoing image compression method when executing a program stored in the memory. Among them, the image compression method includes:

Obtain the image to be compressed;

The first convolutional neural network model is used to perform convolution filtering and compression on the compressed image to obtain network compressed data. Among them, the first convolutional neural network model and the second convolutional neural network model are obtained by training using a preset training set. Including multiple sample images, the second convolutional neural network model is used to decompress the network compressed data by convolution filtering to obtain a decompressed image;

Storage network compresses data.

An embodiment of the present application provides an electronic device for training a CNN model by using multiple images below a resolution threshold and multiple images above a resolution threshold. The CNN model is used to perform convolution filtering and compression on the image to obtain network compression data with a smaller resolution to reduce the occupied storage space. Then, the CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image. Since the CNN model is trained using multiple images below the resolution threshold and multiple images above the resolution threshold, the CNN model can be used to obtain a higher resolution decompressed image, and even the resolution of the decompressed image The rate is higher than the resolution of the original image. This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

Based on the foregoing embodiment of the image decompression method, an embodiment of the present application further provides an electronic device including a processor, a communication interface, a memory, and a communication bus. The processor, the communication interface, and the memory always perform communication with each other through communication. ,

Memory for storing computer programs;

The processor is configured to implement the foregoing image decompression method when executing a program stored in the memory. The image decompression method includes:

Obtain the network compressed data of the target image;

Use the second convolutional neural network model to decompress the network compressed data to obtain a decompressed image; wherein the second convolutional neural network model and the first convolutional neural network model are obtained by training using a preset training set, The training set includes multiple sample images, and the first convolutional neural network model is used to perform convolution filter compression on the compressed image compression to obtain network compressed data;

Output decompressed image.

An embodiment of the present application provides an electronic device for training a CNN model by using multiple images below a resolution threshold and multiple images above a resolution threshold. The CNN model is used to perform convolution filtering and compression on the image to obtain network compression data with a smaller resolution to reduce the occupied storage space. Then, the CNN model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image. Since the CNN model is trained using multiple images below the resolution threshold and multiple images above the resolution threshold, the CNN model can be used to obtain a higher resolution decompressed image, and even the resolution of the decompressed image The rate is higher than the resolution of the original image. This effectively improves the resolution of the image viewed by the user when viewing the compressed image, and improves the user experience.

The above communication bus may be a PCI (Peripheral Component Interconnect, Peripheral Component Interconnect Standard) bus or an EISA (Extended Industry Standard Architecture, Extended Industry Standard Architecture) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, and the like. ,

The communication interface is used for communication between the electronic device and other devices.

The foregoing memory may include RAM (Random Access Memory, Random Access Memory), and may also include NVM (Non-Volatile Memory, non-volatile memory), such as at least one disk memory. Optionally, the memory may also be at least one storage device located far from the foregoing processor.

The above processor may be a general-purpose processor, including a CPU (Central Processing Unit), a NP (Network Processor), etc .; it may also be a DSP (Digital Signal Processing), an ASIC (Application Specific Integrated Circuit (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

Based on the embodiment of the image compression method described above, the embodiment of the present application further provides a machine-readable storage medium. A computer program is stored in the machine-readable storage medium. When the computer program is executed by a processor, any one of the foregoing image compression methods is implemented.

Based on the embodiment of the image decompression method described above, the embodiment of the present application further provides a machine-readable storage medium. A computer program is stored in the machine-readable storage medium. When the computer program is executed by a processor, any of the foregoing image decompression methods is implemented .

Based on the foregoing embodiment of the image compression method, an embodiment of the present application further provides a computer program that implements any one of the foregoing image compression methods when the computer program is executed by a processor.

Based on the foregoing embodiment of the image decompression method, an embodiment of the present application further provides a computer program that implements any one of the foregoing image decompression methods when the computer program is executed by a processor.

It should be noted that in this article, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that these entities or operations There is any such actual relationship or order among them. Moreover, the terms "including", "comprising", or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or device that includes a series of elements includes not only those elements but also those that are not explicitly listed Or other elements inherent to such a process, method, article, or device. Without more restrictions, the elements defined by the sentence "including a ..." do not exclude the existence of other identical elements in the process, method, article, or equipment including the elements.

Each embodiment in this specification is described in a related manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, the image compression device embodiment, the image decompression device embodiment, the electronic device embodiment, the machine-readable storage medium embodiment, and the computer program embodiment are basically similar to the image compression method and the image decompression method. For example, the description is relatively simple. For related points, refer to the description of the image compression method and image decompression method embodiments.

The above descriptions are merely preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Any modification, equivalent replacement, and improvement made within the spirit and principle of this application are included in the protection scope of this application.

Claims (28)

An image compression method, wherein the method includes: Obtain the image to be compressed; The first convolutional neural network model is used to perform convolution filtering and compression on the image to be compressed, to obtain network compressed data; wherein the first convolutional neural network model and the second convolutional neural network model use a preset training set. Obtained through training, the training set includes multiple low-resolution images with a resolution lower than a resolution threshold, and images with a resolution higher than the resolution threshold corresponding to the low-resolution images, and the second convolutional neural network The model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; The network compressed data is stored. The method according to claim 1, further comprising: After obtaining the image to be compressed, performing size compression on the image to be compressed to obtain a size-compressed image; After the network compressed data is obtained, the correspondence between the size compressed image and a preset identifier is stored; the preset identifier is used to indicate the presence of the network compressed data of the image to be compressed. The method according to claim 1 or 2, wherein the image to be compressed is located in a PDF document; or the image to be compressed is located in a Word document; or the image to be compressed is located in an Excel document; or The image to be compressed is located in a WPS document. An image decompression method, wherein the method includes: Obtain the network compressed data of the target image; The second convolutional neural network model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; wherein the second convolutional neural network model and the first convolutional neural network model use preset training. Obtained from training set, the training set includes multiple low-resolution images with a resolution lower than a resolution threshold, and images corresponding to a low-resolution image with a resolution higher than the resolution threshold, the first convolutional nerve The network model is used to perform convolution filter compression on the image to be compressed to obtain the network compression data; The decompressed image is output. The method according to claim 4, wherein the step of obtaining network compressed data of a target image comprises: Obtain the size-compressed image of the target image; Determining whether a preset identifier corresponding to the compressed image of the size is stored; the preset identifier is used to indicate that network compression data of the target image exists; If stored, the network compressed data is obtained. The method according to claim 5, further comprising: If the preset identifier corresponding to the size-compressed image is not stored, the size-compressed image is output. An image compression device, wherein the device includes: An acquisition module for acquiring an image to be compressed; A first compression module, configured to perform convolution filtering and compression on the image to be compressed using a first convolutional neural network model to obtain network compressed data; wherein the first convolutional neural network model and the second convolutional neural network The model is obtained by training using a preset training set, where the training set includes multiple low-resolution images with a resolution lower than the resolution threshold and images with a resolution higher than the resolution threshold corresponding to the low-resolution image. The second convolutional neural network model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; The first storage module is configured to store the network compressed data. The apparatus according to claim 7, further comprising: A second compression module, configured to compress the image to be compressed after obtaining the image to be compressed to obtain a size-compressed image; A second storage module is configured to store the correspondence between the size compressed image and a preset identifier after the network compressed data is obtained, where the preset identifier is used to indicate the existence of network compressed data of the image to be compressed. The device method according to claim 7 or 8, wherein the image to be compressed is located in a PDF document; or the image to be compressed is located in a Word document; or the image to be compressed is located in an Excel document; or The image to be compressed is described in a WPS document. An image decompression device, wherein the device includes: An acquisition module for acquiring network compressed data of a target image; A decompression module, configured to perform convolution filtering and decompression on the network compressed data using a second convolutional neural network model to obtain a decompressed image; wherein the second convolutional neural network model and the first convolutional neural network The model is obtained by training using a preset training set, where the training set includes multiple low-resolution images with a resolution lower than the resolution threshold and images with a resolution higher than the resolution threshold corresponding to the low-resolution image. The first convolutional neural network model is used to perform convolution filter compression on the image to be compressed to obtain the network compression data; An output module, configured to output the decompressed image. The device according to claim 10, wherein the obtaining module is specifically configured to obtain a size-compressed image of the target image; determine whether a preset identifier corresponding to the size-compressed image is stored; Indicating that the network compressed data of the target image exists; if stored, obtaining the network compressed data. The device according to claim 11, wherein the output module is further configured to output the size-compressed image if the preset identifier corresponding to the size-compressed image is not stored. An image compression method, wherein the method includes: Obtain the image to be compressed; The first convolutional neural network model is used to perform convolution filtering and compression on the image to be compressed, to obtain network compressed data; wherein the first convolutional neural network model and the second convolutional neural network model use a preset training set. Obtained through training, the training set includes multiple sample images, and the second convolutional neural network model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; The network compressed data is stored. An image decompression method, wherein the method includes: Obtain the network compressed data of the target image; The second convolutional neural network model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; wherein the second convolutional neural network model and the first convolutional neural network model use preset training. The training set is obtained, the training set includes multiple sample images, and the first convolutional neural network model is used to perform convolution filter compression on the compressed image to be compressed to obtain the network compressed data; The decompressed image is output. An image compression device, wherein the device includes: An acquisition module for acquiring an image to be compressed; A first compression module, configured to perform convolution filtering and compression on the image to be compressed using a first convolutional neural network model to obtain network compressed data; wherein the first convolutional neural network model and the second convolutional neural network The model is obtained by training using a preset training set, where the training set includes multiple sample images, and the second convolutional neural network model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; The first storage module is configured to store the network compressed data. An image decompression device, wherein the device includes: An acquisition module for acquiring network compressed data of a target image; A decompression module, configured to perform convolution filtering and decompression on the network compressed data using a second convolutional neural network model to obtain a decompressed image; wherein the second convolutional neural network model and the first convolutional neural network The model is obtained by training using a preset training set, where the training set includes multiple sample images, and the first convolutional neural network model is used to perform convolution filtering and compression compression on a compressed image to obtain the network compressed data; An output module, configured to output the decompressed image. An electronic device, comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete communication with each other through the communication bus; The memory is used to store a computer program; The processor is configured to execute a program stored in the memory to implement: Obtain the image to be compressed; The first convolutional neural network model is used to perform convolution filtering and compression on the image to be compressed, to obtain network compressed data; wherein the first convolutional neural network model and the second convolutional neural network model use a preset training set. Obtained through training, the training set includes multiple low-resolution images with a resolution lower than a resolution threshold, and images with a resolution higher than the resolution threshold corresponding to the low-resolution images, and the second convolutional neural network The model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; The network compressed data is stored. An electronic device, comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete communication with each other through the communication bus; The memory is used to store a computer program; The processor is configured to execute a program stored in the memory to implement: Obtain the network compressed data of the target image; The second convolutional neural network model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; wherein the second convolutional neural network model and the first convolutional neural network model use preset training. Obtained from training set, the training set includes multiple low-resolution images with a resolution lower than a resolution threshold, and images corresponding to a low-resolution image with a resolution higher than the resolution threshold, the first convolutional nerve The network model is used to perform convolution filter compression on the image to be compressed to obtain the network compression data; The decompressed image is output. An electronic device, comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete communication with each other through the communication bus; The memory is used to store a computer program; The processor is configured to execute a program stored in the memory to implement: Obtain the image to be compressed; The first convolutional neural network model is used to perform convolution filtering and compression on the image to be compressed, to obtain network compressed data; wherein the first convolutional neural network model and the second convolutional neural network model use a preset training set. Obtained through training, the training set includes multiple sample images, and the second convolutional neural network model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; The network compressed data is stored. An electronic device, comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete communication with each other through the communication bus; The memory is used to store a computer program; The processor is configured to execute a program stored in the memory to implement: Obtain the network compressed data of the target image; The second convolutional neural network model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; wherein the second convolutional neural network model and the first convolutional neural network model use preset training. The training set is obtained, the training set includes multiple sample images, and the first convolutional neural network model is used to perform convolution filter compression on the compressed image to be compressed to obtain the network compressed data; The decompressed image is output. A machine-readable storage medium, characterized in that a computer program is stored in the machine-readable storage medium, and the computer program is implemented when executed by a processor: Obtain the image to be compressed; The first convolutional neural network model is used to perform convolution filtering and compression on the image to be compressed, to obtain network compressed data; wherein the first convolutional neural network model and the second convolutional neural network model use a preset training set. Obtained through training, the training set includes multiple low-resolution images with a resolution lower than a resolution threshold, and images with a resolution higher than the resolution threshold corresponding to the low-resolution images, and the second convolutional neural network The model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; The network compressed data is stored. A machine-readable storage medium, characterized in that a computer program is stored in the machine-readable storage medium, and the computer program is implemented when executed by a processor: Obtain the network compressed data of the target image; The second convolutional neural network model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; wherein the second convolutional neural network model and the first convolutional neural network model use preset training. Obtained from training set, the training set includes multiple low-resolution images with a resolution lower than a resolution threshold, and images corresponding to a low-resolution image with a resolution higher than the resolution threshold, the first convolutional nerve The network model is used to perform convolution filter compression on the image to be compressed to obtain the network compression data; The decompressed image is output. A machine-readable storage medium, characterized in that a computer program is stored in the machine-readable storage medium, and the computer program is implemented when executed by a processor: Obtain the image to be compressed; The first convolutional neural network model is used to perform convolution filtering and compression on the image to be compressed, to obtain network compressed data; wherein the first convolutional neural network model and the second convolutional neural network model use a preset training set. Obtained through training, the training set includes multiple sample images, and the second convolutional neural network model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; The network compressed data is stored. A machine-readable storage medium, characterized in that a computer program is stored in the machine-readable storage medium, and the computer program is implemented when executed by a processor: Obtain the network compressed data of the target image; Convolutional filtering and decompression of the network compressed data using a second convolutional neural network model to obtain a decompressed image; wherein the second convolutional neural network model and the first convolutional neural network model use preset training The training set is obtained, the training set includes multiple sample images, and the first convolutional neural network model is used to perform convolution filter compression on the compressed image to be compressed to obtain the network compressed data; The decompressed image is output. A computer program, characterized in that the computer program is implemented when executed by a processor: Obtain the image to be compressed; The first convolutional neural network model is used to perform convolution filtering and compression on the image to be compressed, to obtain network compressed data; wherein the first convolutional neural network model and the second convolutional neural network model use a preset training set. Obtained through training, the training set includes multiple low-resolution images with a resolution lower than a resolution threshold, and images with a resolution higher than the resolution threshold corresponding to the low-resolution images, and the second convolutional neural network The model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; The network compressed data is stored. A computer program, characterized in that the computer program is implemented when executed by a processor: Obtain the network compressed data of the target image; The second convolutional neural network model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; wherein the second convolutional neural network model and the first convolutional neural network model use preset training. Obtained from training set, the training set includes multiple low-resolution images with a resolution lower than a resolution threshold, and images corresponding to a low-resolution image with a resolution higher than the resolution threshold, the first convolutional nerve The network model is used to perform convolution filter compression on the image to be compressed to obtain the network compression data; The decompressed image is output. A computer program, characterized in that the computer program is implemented when executed by a processor: Obtain the image to be compressed; The first convolutional neural network model is used to perform convolution filtering and compression on the image to be compressed, to obtain network compressed data; wherein the first convolutional neural network model and the second convolutional neural network model use a preset training set. Obtained through training, the training set includes multiple sample images, and the second convolutional neural network model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; The network compressed data is stored. A computer program, characterized in that the computer program is implemented when executed by a processor: Obtain the network compressed data of the target image; The second convolutional neural network model is used to perform convolution filtering and decompression on the network compressed data to obtain a decompressed image; wherein the second convolutional neural network model and the first convolutional neural network model use preset training. The training set is obtained, the training set includes multiple sample images, and the first convolutional neural network model is used to perform convolution filter compression on the compressed image to be compressed to obtain the network compressed data; The decompressed image is output.

Images