CN108717542B - Method, apparatus and computer-readable storage medium for identifying text region - Google Patents

Abstract

The present disclosure relates to a method, apparatus, and computer-readable storage medium for identifying text regions. Using this method, firstly, the feature information of the image to be recognized is input into the text area recognition model to obtain the first probability and the second probability of each first image area in the to-be-recognized image, and then, according to each first image The first probability of the area is to filter out the image area containing the text, and then, on the basis of the screened image area containing the text, the second probability of the image area is further analyzed to determine whether the image area is adjacent to it. The image areas of the image are combined, and finally, according to the combined image area, the text area in the image to be recognized is determined. Therefore, by determining the first probability and the second probability of each first image region in the image to be recognized, the text region can be directly calculated, and a text region recognition method with high accuracy and fast recognition speed is provided.

Description

Method, apparatus and computer-readable storage medium for identifying text region

technical field

The present disclosure relates to the field of image processing, and in particular, to a method, a device, and a computer-readable storage medium for recognizing a text area.

Background technique

With the rapid development of Internet technology, there are more and more pictures on the Internet, which causes inconvenience for users to obtain desired pictures. Since pictures not only contain pictures, but also some text, people tend to filter out the pictures they want by searching for the text in the pictures. Driven by this trend, OCR (Optical Character Recognition, Optical Character Recognition) technology came into being. pregnancy.

OCR refers to the technology of recognizing optical characters through image processing and pattern recognition technology. It generally includes two stages: text area recognition and text recognition. Among them, text area recognition is used to identify the position of the text in the image. Text recognition is used to recognize the text in the text area.

Under normal circumstances, most of the recognition text areas are based on the Adaboost method. This method mainly uses artificially designed features to identify text areas. Due to the error of artificially designed features, the accuracy of using this method for text area recognition is not high. .

SUMMARY OF THE INVENTION

In order to overcome the problems existing in the related art, the present disclosure provides a method, an apparatus and a computer-readable storage medium for recognizing a text area.

According to a first aspect of the embodiments of the present disclosure, there is provided a method for recognizing a text area, including:

Input the feature information of the image to be recognized into the text area recognition model, and obtain the first probability and the second probability of each first image area in the to-be-recognized image, where the first probability indicates that the first image area is a text area The second probability represents the probability that the first image area is connected to an adjacent image area, wherein the text area recognition model is used to identify the text area in the image;

determining the first image area with a first probability greater than a first probability threshold as a second image area;

merging two adjacent second image regions whose second probability is greater than a second probability threshold;

According to the combined image area, determine the text area in the to-be-recognized image.

Optionally, after determining the first image area with a first probability greater than a first probability threshold as the second image area, the method further includes:

Inputting the second image area into the text area recognition model to obtain the position offset of each of the second image areas in the to-be-recognized image;

adjusting the position of the second image area according to the position offset of each of the second image areas;

Merging the two adjacent second image regions whose second probability is greater than the second probability threshold, including:

Merging two adjacent and position-adjusted second image regions whose second probability is greater than the second probability threshold.

Optionally, determining the text area in the to-be-recognized image according to the merged image area includes:

Determine the smallest bounding rectangle of the merged image area;

The area where the minimum circumscribed rectangle is located is determined as the text area in the to-be-recognized image.

Optionally, the method further includes:

According to the feature information of the sample image and the text area in the sample image, the convolutional neural network is trained to obtain the text area recognition model.

According to a second aspect of the embodiments of the present disclosure, there is provided an apparatus for recognizing a text area, including:

The probability obtaining module is configured to input the feature information of the image to be recognized into the text region recognition model, and obtain the first probability and the second probability of each first image region in the to-be-recognized image, the first probability representing the The probability that the first image region is a text region, the second probability represents the probability that the image region is connected to an adjacent image region, wherein the text region recognition model is used to identify the text region in the image;

a first determination module, configured to determine the first image area with a first probability greater than a first probability threshold as a second image area;

a merging module, configured to merge two adjacent second image regions whose second probability is greater than a second probability threshold;

The second determination module is configured to determine the text area in the image to be recognized according to the merged image area.

Optionally, the device further includes:

an offset obtaining module, configured to input the second image area into the text area recognition model to obtain a position offset of each of the second image areas in the to-be-recognized image;

an adjustment module, configured to adjust the position of the second image area according to the position offset of each of the second image areas;

The merging module includes:

The merging sub-module is configured to merge two adjacent and position-adjusted second image regions whose second probability is greater than the second probability threshold.

Optionally, the merging module includes:

a first determination submodule, configured to determine the minimum circumscribed rectangle of the merged image area;

The second determination submodule is configured to determine the area where the minimum circumscribed rectangle is located as the text area in the image to be recognized.

Optionally, the device further includes:

The training module is configured to train the convolutional neural network according to the feature information of the sample image and the text area in the sample image to obtain the text area recognition model.

According to a third aspect of the embodiments of the present disclosure, there is provided a computer-readable storage medium on which computer program instructions are stored, and when the program instructions are executed by a processor, implement the method for identifying a text region provided by the first aspect of the present disclosure. step.

In the embodiment of the present disclosure, first, the feature information of the image to be recognized is input into the text area recognition model to obtain the first probability and the second probability of each first image area in the to-be-recognized image, and then, according to each The first probability of the first image area is to filter out the image area containing the text, and then, on the basis of the screened out image area containing the text, the second probability of the image area is further analyzed to determine whether the image area It is combined with its adjacent image areas, and finally, according to the combined image areas, the text area in the image to be recognized is determined. Therefore, by determining the first probability and the second probability of each first image region in the image to be recognized, the text region can be directly calculated, and a text region recognition method with high accuracy and fast recognition speed is provided.

It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description serve to explain the principles of the disclosure.

Fig. 1 is a flow chart of a method for recognizing a text area according to an exemplary embodiment.

Fig. 2 is another flowchart of a method for recognizing a text area according to an exemplary embodiment.

Fig. 3 is another flowchart of a method for recognizing a text area according to an exemplary embodiment.

Fig. 4 is another flowchart of a method for recognizing a text area according to an exemplary embodiment.

Fig. 5 is a block diagram of an apparatus for recognizing a text area according to an exemplary embodiment.

Fig. 6 is another block diagram of an apparatus for recognizing a text area according to an exemplary embodiment.

Fig. 7 is a block diagram of a merging module in an apparatus for recognizing text regions according to an exemplary embodiment.

Fig. 8 is another block diagram of an apparatus for recognizing a text area according to an exemplary embodiment.

Fig. 9 is a block diagram of an apparatus for recognizing a text area according to an exemplary embodiment.

Detailed ways

Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. Where the following description refers to the drawings, the same numerals in different drawings refer to the same or similar elements unless otherwise indicated. The implementations described in the illustrative examples below are not intended to represent all implementations consistent with this disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as recited in the appended claims.

An embodiment of the present disclosure provides a method for identifying a text area. Please refer to FIG. 1 , which is a flowchart of a method for recognizing a text area according to an exemplary embodiment. As shown in FIG. 1 , the method for recognizing a text area provided by an embodiment of the present disclosure includes the following steps.

In step S11, the feature information of the image to be recognized is input into the text area recognition model, and the first probability and the second probability of each first image area in the image to be recognized are obtained, and the first probability represents the first probability The probability that an image region is a text region, the second probability represents the probability that the first image region is connected to an adjacent image region, wherein the text region recognition model is used to identify text regions in an image.

In step S12, the first image area with a first probability greater than a first probability threshold is determined as a second image area.

In step S13, two adjacent second image regions whose second probability is greater than a second probability threshold are merged.

In step S14, the character area in the image to be recognized is determined according to the merged image area.

CNN (convolutional neural network; convolutional neural network) is a feedforward neural network, which is mainly composed of a convolution layer and a connection layer. The connection layer has requirements on the size of the input image, and in practical applications, Different images may have different sizes, which cannot meet the requirements of the connection layer in the CNN, which may cause the CNN to fail to recognize the image. Meet the requirements of the connection layer in this CNN.

In the field of image processing, since the pixel value of each pixel in the image can only represent the basic primitive color and its gray level, there is no high-level semantic information. The high-level semantic information is used to indicate what is in the image The object, where the object is located, etc. Therefore, after the image to be recognized is zoomed, it is not possible to directly recognize the text area, but to obtain high-level semantic information of the image. Specifically, the zoomed image to be recognized can be input into the CNN to obtain a Corresponding to the feature matrix of the image (hereinafter referred to as the feature map), the high-level semantic information of the image to be recognized is identified from the feature map, and then the region where the text is located can be further identified.

In the embodiment of the present disclosure, feature information of the image to be recognized may be extracted from a feature map of the image to be recognized, and the feature information includes high-level semantic information of the image to be recognized. First, in step S11, the feature information of the image to be recognized is input into the text area recognition model to obtain the first probability and the second probability of each first image area in the to-be-recognized image, and the first probability of each first image area is obtained. The probability represents the probability that the first image area is a text area, and the second probability of each first image area represents the probability of whether the first image area is connected to an adjacent image area. The first image area is any image area in the image to be recognized.

Specifically, a plurality of anchor points are preset in the image to be recognized, and some of the anchor points are used to represent the area in the image, referred to as area anchor points for short, and the other part of the anchor points are used to indicate whether the adjacent areas in the image are not. A connection is required, referred to as a connection anchor for short. In the original image, multiple rectangular areas with rotation are preset, and the image to be recognized is divided into multiple image areas. The rectangular area with rotation can be expressed as (x, y, w, h, θ), and x, y is The coordinates of the center point of the rectangular area with rotation, w, h are the width and height of the rectangular area with rotation, θ is the rotation angle of the rectangular area with rotation relative to the horizontal direction, and each area anchor point corresponds to a belt The rotated rectangular area, that is, corresponds to an image area in the image to be recognized. Optionally, the preset rectangular regions with rotation may overlap each other.

After the feature information of the image to be recognized is input into the text area recognition model, the text area recognition model recognizes each first image area in the to-be-recognized image, and the recognition result of each first image area includes the first image area. a probability and a second probability, wherein the first probability is used to represent the probability that the first image area is a text area, and the second probability is used to represent the probability that the first image area is connected to an adjacent image area.

Optionally, the text region recognition model in step S11 is a model for recognizing text regions in an image, and can be obtained by training a fully convolutional network. Specifically, the text region recognition model can be obtained by executing step S17. Fig. 2 is another flowchart of a method for recognizing a text area according to an exemplary embodiment. As shown in FIG. 2, the method further includes step S17 in addition to steps S11-S14.

In step S17, the convolutional neural network is trained according to the feature information of the sample image and the text area in the sample image, so as to obtain the text area recognition model.

Usually, the coefficients in the convolutional neural network are randomly generated. Using the randomly generated coefficients, the accuracy of identifying the text area of the image to be recognized cannot be guaranteed. Therefore, when using the convolutional neural network for text area recognition Previously, the convolutional network could be trained according to the user's demand for text recognition accuracy to adjust the coefficients in the convolutional neural network, so that the text region recognized by the convolutional neural network is more accurate.

Specifically, the sample image to be identified can be input into the convolutional neural network, and the resultant sample image can be output through the processing of the convolutional neural network. The result sample image has at least one rectangular area, and the rectangular area is used to represent the identified The text area, since the rectangular area is identified based on randomly generated coefficients, it is possible that the rectangular area cannot accurately cover the text area in the sample image to be recognized. Identify the sample image, the area enclosed by the rectangular frame is the text area in the sample image to be recognized) and compare, and adjust the coefficients in the convolutional neural network according to the error between the two to reduce the error between the result sample image and the target sample image. . Repeat the above method several times until the error between the result sample image output by the convolutional neural network and the target sample image meets the preset requirements, and the convolutional neural network after the coefficient adjustment is completed is the text area recognition model, wherein the preset requirements It is preset according to the requirement of text recognition accuracy.

After step S11 is performed, step S12 is performed, and each first image area in the to-be-recognized image is screened according to the first probability of each first image area in the to-be-recognized image determined in step S11. Specifically, in the recognition result output by the text region recognition model, if the value of the first probability of the first image region is large, it can be considered that the first image region contains text, and the value of the first probability is small, it is considered that the value of the first probability is small. The first image area does not contain text, therefore, a first probability threshold can be preset and compared with the value of the first probability of each first image area, and the value of the first probability greater than the first probability threshold The corresponding first image area is determined as an area containing text, and the determined area containing text is selected as the second image area. Optionally, the first image area corresponding to the value of the first probability smaller than the first probability threshold is determined as an area that does not contain text, and no processing is performed on the area.

The first probability threshold is used to represent that when the value of the first probability is greater than the first probability threshold, the first image area corresponding to the value of the first probability contains text, and when the value of the first probability is less than the first probability When the threshold is set, the first image area corresponding to the value of the first probability does not contain text.

The image area (that is, the second image area) screened in step S12 can only represent that the area contains text, but the size of the text is uncertain, and the preset rectangular area is determined, and the rectangular area is not necessarily The text is completely contained, that is, the filtered image area may contain only a part of the text, and the image area adjacent to the image area contains the remaining part of the text. Therefore, in the embodiment of the present disclosure, it is necessary to further analyze whether the image area only contains a part of the text by performing step S13 on the basis of the screened image area containing the text.

In step S13, the second probability is further analyzed for the image regions containing text that are screened out in step S12, and the screened image regions are merged according to the second probability. Specifically, in the recognition result output by the text region recognition model, the second probability of each first image region includes a plurality of probability values, and the plurality of probability values respectively represent a plurality of probability values adjacent to the first image region Whether the image area contains the remaining part of the text, where the probability value is larger, it can be considered that the adjacent image area corresponding to the probability value also contains a part of the text, the two are connected, and the probability value is higher. If it is small, it can be considered that the adjacent image area corresponding to the probability value does not contain a part of the text, and the two are not connected. Therefore, a second probability threshold can be preset, compared with the value of the second probability of the image region screened in step S12, and the adjacent image regions corresponding to the probability value greater than the second probability threshold are determined as containing A region with a part of the text, and merge this region with the filtered image region.

The above steps are performed for each image area containing text screened out in step S12, and finally, in step S14, the combined image area can be determined as the text area in the image to be recognized.

In the embodiment of the present disclosure, first, the feature information of the image to be recognized is input into the text area recognition model to obtain the first probability and the second probability of each first image area in the to-be-recognized image, and then, according to each The first probability of the first image area is to filter out the image area containing the text, and then, on the basis of the screened out image area containing the text, the second probability of the image area is further analyzed to determine whether the image area It is combined with its adjacent image areas, and finally, according to the combined image areas, the text area in the image to be recognized is determined. Therefore, by determining the first probability and the second probability of each first image region in the image to be recognized, the text region can be directly calculated, and a text region recognition method with high accuracy and fast recognition speed is provided.

Optionally, FIG. 3 is another flowchart of a method for recognizing a text area according to an exemplary embodiment. As shown in FIG. 3, after step S12, the method further includes the following steps.

In step S15, the second image area is input into the text area recognition model to obtain the position offset of each of the second image areas in the image to be recognized.

In step S16, the position of the second image area is adjusted according to the position offset of each of the second image areas.

Correspondingly, step S13 specifically includes step S131.

In step S131, two adjacent and position-adjusted second image regions whose second probability is greater than the second probability threshold are merged.

In the embodiment of the present disclosure, the sizes and positions in the image of the plurality of preset rectangular regions with rotation are determined, and the size of each first image region into which the to-be-recognized image is divided is also determined. In the image of , the position of the text is not the same. Therefore, the image area screened out in step S12 may contain a small area of the text part. At this time, just adjust the position of the image area slightly to make The area of the image area containing the text portion is increased, which facilitates the rapid identification of the text area. Therefore, after the second image area is obtained in step S12, it needs to be input into the text area recognition model to obtain the position offset (Δx, Δy, Δw, Δh, Δθ) of the second image area, and according to the Position offset, adjust the position of the second image area, so that the larger area in the image area can contain text, wherein, the position of the adjusted image area is (x+Δx,y+Δy , w+Δw, h+Δh, θ+Δθ).

After the image area is adjusted, the position-adjusted image areas are merged according to the second probability of the position-adjusted image area. The specific implementation manner is as described above, and will not be repeated here.

With the above technical solution, considering that the area of the image area containing text may be small, before judging whether the image area is merged with the adjacent image area, first fine-tune the image area, and increase the size of the image area in the image area. Covers the area of the text, reduces the number of operations for region merging, and further improves the speed of text region recognition.

Optionally, FIG. 4 is another flowchart of a method for recognizing a text area according to an exemplary embodiment. As shown in FIG. 4 , step S14 in FIG. 1 specifically includes the following steps.

In step S141, the minimum circumscribed rectangle of the merged image area is determined.

In step S142, the area where the minimum circumscribed rectangle is located is determined as the text area in the image to be recognized.

Usually, since the adjacent image areas of the image area are located in the eight directions of the upper, lower, left, rear, front left, rear left, front right and rear right of the image area, the image area can be the same as the above If the image areas in any of the eight directions are merged, the merged text area may not be regular. In the text recognition technology, a rectangle is usually used to represent the text area. Therefore, after the image areas are merged, In order to ensure that the text is in the same rectangular frame, the combined image area needs to be processed to obtain a rectangular area.

Specifically, first, according to the position of the merged image area, the minimum circumscribed rectangle of the merged graphic area is determined, and the minimum circumscribed rectangle contains the text in the image to be recognized, and then the area where the minimum circumscribed rectangle is located is determined The text area in the image to be recognized.

Based on the same inventive concept, an embodiment of the present disclosure also provides an apparatus for recognizing a text area. Fig. 5 is a block diagram of an apparatus for recognizing a text area according to an exemplary embodiment. 5, the apparatus 500 includes:

The probability obtaining module 501 is configured to input the feature information of the image to be recognized into the text region recognition model, and obtain the first probability and the second probability of each first image region in the to-be-recognized image, and the first probability represents the the probability that the first image region is a text region, the second probability represents the probability that the image region is connected to an adjacent image region, wherein the text region recognition model is used to identify the text region in the image;

a first determination module 502, configured to determine the first image area with a first probability greater than a first probability threshold as a second image area;

Merging module 503, configured to merge two adjacent second image regions whose second probability is greater than a second probability threshold;

The second determination module 504 is configured to determine the text area in the image to be recognized according to the combined image area.

Optionally, FIG. 6 is another block diagram of an apparatus for recognizing a text area according to an exemplary embodiment. As shown in FIG. 6, the apparatus 500 further includes:

an offset obtaining module 505, configured to input the second image region into the text region recognition model to obtain the position offset of each of the second image regions in the to-be-recognized image;

The adjustment module 506 is configured to adjust the position of the second image area according to the position offset of each of the second image areas;

The merging module 503 includes:

The merging sub-module 5031 is configured to merge two adjacent and position-adjusted second image regions whose second probability is greater than the second probability threshold.

Optionally, FIG. 7 is a block diagram of a merging module in an apparatus for recognizing text regions according to an exemplary embodiment. As shown in Figure 7, the merging module 503 includes:

The first determination sub-module 5032 is configured to determine the minimum circumscribed rectangle of the merged image area;

The second determination sub-module 5033 is configured to determine the area where the minimum circumscribed rectangle is located as the text area in the image to be recognized.

Optionally, FIG. 8 is another block diagram of an apparatus for recognizing a text area according to an exemplary embodiment. As shown in FIG. 8, the apparatus 500 further includes:

The training module 507 is configured to train the convolutional neural network according to the feature information of the sample image and the text area in the sample image, so as to obtain the text area recognition model.

Regarding the apparatus in the above-mentioned embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be described in detail here.

The present disclosure also provides a computer-readable storage medium on which computer program instructions are stored, and when the program instructions are executed by a processor, implement the steps of the method for recognizing a text area provided by the present disclosure.

FIG. 9 is a block diagram of an apparatus 800 for recognizing text regions according to an exemplary embodiment. For example, apparatus 800 may be a mobile phone, computer, digital broadcast terminal, messaging device, game console, tablet device, medical device, fitness device, personal digital assistant, and the like.

9, the apparatus 800 may include one or more of the following components: a processing component 802, a memory 804, a power component 806, a multimedia component 808, an audio component 810, an input/output (I/O) interface 812, a sensor component 814, and communication component 816.

The processing component 802 generally controls the overall operation of the device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to perform all or part of the steps of the above-described method of identifying text regions. Additionally, processing component 802 may include one or more modules that facilitate interaction between processing component 802 and other components. For example, processing component 802 may include a multimedia module to facilitate interaction between multimedia component 808 and processing component 802.

Memory 804 is configured to store various types of data to support operations at device 800 . Examples of such data include instructions for any application or method operating on device 800, contact data, phonebook data, messages, pictures, videos, and the like. Memory 804 may be implemented by any type of volatile or nonvolatile storage device or combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Magnetic or Optical Disk.

Power component 806 provides power to various components of device 800 . Power components 806 may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power to device 800 .

Multimedia component 808 includes a screen that provides an output interface between the device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touch, swipe, and gestures on the touch panel. The touch sensor may not only sense the boundaries of a touch or swipe action, but also detect the duration and pressure associated with the touch or swipe action. In some embodiments, the multimedia component 808 includes a front-facing camera and/or a rear-facing camera. When the apparatus 800 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera may receive external multimedia data. Each of the front and rear cameras can be a fixed optical lens system or have focal length and optical zoom capability.

Audio component 810 is configured to output and/or input audio signals. For example, audio component 810 includes a microphone (MIC) that is configured to receive external audio signals when device 800 is in operating modes, such as call mode, recording mode, and voice recognition mode. The received audio signal may be further stored in memory 804 or transmitted via communication component 816 . In some embodiments, audio component 810 also includes a speaker for outputting audio signals.

The I/O interface 812 provides an interface between the processing component 802 and a peripheral interface module, which may be a keyboard, a click wheel, a button, or the like. These buttons may include, but are not limited to: home button, volume buttons, start button, and lock button.

Sensor assembly 814 includes one or more sensors for providing status assessment of various aspects of device 800 . For example, the sensor assembly 814 can detect the open/closed state of the device 800, the relative positioning of components, such as the display and keypad of the device 800, and the sensor assembly 814 can also detect a change in the position of the device 800 or a component of the device 800 , the presence or absence of user contact with the device 800 , the orientation or acceleration/deceleration of the device 800 and the temperature change of the device 800 . Sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects in the absence of any physical contact. Sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

Communication component 816 is configured to facilitate wired or wireless communication between apparatus 800 and other devices. Device 800 may access wireless networks based on communication standards, such as WiFi, 2G or 3G, or a combination thereof. In one exemplary embodiment, the communication component 816 receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

In an exemplary embodiment, apparatus 800 may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable A gate array (FPGA), a controller, a microcontroller, a microprocessor or other electronic components are implemented for implementing the above-mentioned method for recognizing a text area.

In an exemplary embodiment, there is also provided a non-transitory computer-readable storage medium including instructions, such as a memory 804 including instructions, executable by the processor 820 of the apparatus 800 to accomplish the above-described method of recognizing a text region. For example, the non-transitory computer-readable storage medium may be ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like.

Other embodiments of the present disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the present disclosure. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or techniques in the technical field not disclosed by the present disclosure . The specification and examples are to be regarded as exemplary only, with the true scope and spirit of the disclosure being indicated by the following claims.

It is to be understood that the present disclosure is not limited to the precise structures described above and illustrated in the accompanying drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims (8)

1. A method for identifying text regions, comprising:

inputting feature information of an image to be recognized into a character region recognition model, and obtaining a first probability and a second probability of each first image region in the image to be recognized, wherein the first probability represents the probability that the first image region is a character region, and the second probability represents the probability that the first image region and an adjacent image region have associated character content, and the character region recognition model is used for recognizing the character region in the image;

determining the first image region having a first probability greater than a first probability threshold as a second image region;

merging two adjacent second image regions with second probabilities larger than a second probability threshold;

determining a character area in the image to be recognized according to the combined image area;

wherein, after determining the first image region having a first probability greater than a first probability threshold as a second image region, the method further comprises:

inputting the second image area into the character area recognition model to obtain the position offset of each second image area in the image to be recognized;

adjusting the position of each second image area according to the position offset of the second image area;

merging two adjacent second image regions with second probabilities both greater than a second probability threshold, including:

and merging two adjacent second image areas with the second probabilities larger than a second probability threshold value after position adjustment.

2. The method according to claim 1, wherein the determining a text region in the image to be recognized according to the merged image region comprises:

determining the minimum circumscribed rectangle of the combined image area;

and determining the area where the minimum circumscribed rectangle is located as a character area in the image to be recognized.

3. The method of claim 1, further comprising:

and training the convolutional neural network according to the characteristic information of the sample image and the character area in the sample image to obtain the character area recognition model.

4. An apparatus for recognizing a text region, comprising:

a probability obtaining module configured to input feature information of an image to be recognized into a text region recognition model, and obtain a first probability and a second probability of each first image region in the image to be recognized, wherein the first probability represents a probability that the first image region is a text region, and the second probability represents a probability that the first image region and an adjacent image region have associated text content, and the text region recognition model is used for recognizing the text region in the image;

a first determination module configured to determine the first image region having a first probability greater than a first probability threshold as a second image region;

a merging module configured to merge two adjacent second image regions having a second probability that is greater than a second probability threshold;

the second determining module is configured to determine a character area in the image to be recognized according to the combined image area;

wherein the apparatus further comprises:

an offset obtaining module, configured to input the second image area into the character area recognition model, to obtain a position offset of each second image area in the image to be recognized;

an adjusting module configured to adjust a position of each of the second image regions according to a position offset amount of the second image region;

the merging module comprises:

and the merging submodule is configured to merge two adjacent and position-adjusted second image regions of which the second probabilities are both greater than a second probability threshold.

5. The apparatus of claim 4, wherein the merging module comprises:

a first determination submodule configured to determine a minimum bounding rectangle of the merged image region;

and the second determining submodule is configured to determine the area where the minimum circumscribed rectangle is located as a character area in the image to be recognized.

6. The apparatus of claim 4, further comprising:

and the training module is configured to train the convolutional neural network according to the characteristic information of the sample image and the character area in the sample image to obtain the character area recognition model.

7. An apparatus for recognizing a text region, comprising:

a processor;

a memory for storing processor-executable instructions;

wherein the processor is configured to:

inputting feature information of an image to be recognized into a character region recognition model, and obtaining a first probability and a second probability of each first image region in the image to be recognized, wherein the first probability represents the probability that the first image region is a character region, and the second probability represents the probability that the first image region and an adjacent image region have associated character content, and the character region recognition model is used for recognizing the character region in the image;

determining the first image region having a first probability greater than a first probability threshold as a second image region;

merging two adjacent second image regions with second probabilities larger than a second probability threshold;

determining a character area in the image to be recognized according to the combined image area;

wherein, after determining the first image region having a first probability greater than a first probability threshold as a second image region, the processor is further to:

inputting the second image area into the character area recognition model to obtain the position offset of each second image area in the image to be recognized;

adjusting the position of each second image area according to the position offset of the second image area;

merging two adjacent second image regions with second probabilities both greater than a second probability threshold, including:

and merging two adjacent second image areas with the second probabilities larger than a second probability threshold value after position adjustment.

8. A computer-readable storage medium, on which computer program instructions are stored, which program instructions, when executed by a processor, carry out the steps of the method according to any one of claims 1 to 3.

Images