[go: up one dir, main page]

WO2002078350A1 - Procede de presuppression du bruit d'une image - Google Patents

Procede de presuppression du bruit d'une image Download PDF

Info

Publication number
WO2002078350A1
WO2002078350A1 PCT/CN2002/000110 CN0200110W WO02078350A1 WO 2002078350 A1 WO2002078350 A1 WO 2002078350A1 CN 0200110 W CN0200110 W CN 0200110W WO 02078350 A1 WO02078350 A1 WO 02078350A1
Authority
WO
WIPO (PCT)
Prior art keywords
threshold
image
dct coefficient
value
filter
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2002/000110
Other languages
English (en)
French (fr)
Inventor
Lianhuan Xiong
Jing Wang
Zhen Chen
Hongyuan Wang
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huawei Technologies Co Ltd
Original Assignee
Huawei Technologies Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Huawei Technologies Co Ltd filed Critical Huawei Technologies Co Ltd
Priority to DE60220380T priority Critical patent/DE60220380T2/de
Priority to EP02703466A priority patent/EP1365590B1/en
Publication of WO2002078350A1 publication Critical patent/WO2002078350A1/zh
Priority to US10/601,732 priority patent/US7194030B2/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/85—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/124—Quantisation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/132—Sampling, masking or truncation of coding units, e.g. adaptive resampling, frame skipping, frame interpolation or high-frequency transform coefficient masking
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/136—Incoming video signal characteristics or properties
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/164—Feedback from the receiver or from the transmission channel
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/18—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a set of transform coefficients
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/189—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the adaptation method, adaptation tool or adaptation type used for the adaptive coding
    • H04N19/196—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the adaptation method, adaptation tool or adaptation type used for the adaptive coding being specially adapted for the computation of encoding parameters, e.g. by averaging previously computed encoding parameters
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/80—Details of filtering operations specially adapted for video compression, e.g. for pixel interpolation

Definitions

  • the present invention relates to the field of image communication in multimedia technology, and in particular, to a method for pre-denoising images at a transmitting end. Background of the invention
  • Video images usually contain a large amount of data, which is not convenient for direct video image communication (such as conference television, video phone, distance learning, etc.) and storage (such as multimedia databases, VCD, DVD, etc.), especially for video business systems. That is to say, in addition to video, it also includes audio, T.120 data, control information, and other content, and has high requirements for real-time, interactivity, and image shield volume. Therefore, video and image data are generally compressed at a large ratio. In order to get satisfactory video results.
  • Existing international standards for video image compression include H.261, H.263 formulated by the International Telecommunication Union ITU-T and MPEG1, MPEG2, MPEG4 formulated by the International Organization for Standardization ISO. They respectively contain some of today's important compression technologies, including Many common places, such as: CIF (Common Intermediate Format) format, motion compensation and DCT (Discrete Cosine Transform) mixed mode, etc.
  • the acquired video images generally inevitably contain some unwanted noise, such as high-frequency impulse noise (the luminance and chrominance components If the value changes abruptly) and random noise (such as noise caused by analog-to-digital conversion, quantization noise during signal sampling), if the denoising process is not performed before compression encoding, the compression efficiency of the image will be greatly affected. Therefore, in order to obtain better compression efficiency and image quality, the necessary pre-processing and post-processing are generally performed to remove or reduce the noise in the image.
  • unwanted noise such as high-frequency impulse noise (the luminance and chrominance components If the value changes abruptly) and random noise (such as noise caused by analog-to-digital conversion, quantization noise during signal sampling)
  • linear filtering such as 1-D FIR (-dimensional finite impulse response) filtering, 2-D FIR (two-dimensional finite impulse response) filtering, etc .
  • nonlinear filtering such as median Filtering, threshold filtering, etc.
  • Reference document [1] that is, the method disclosed in US5787203: "Method and system for filtering compressed video images”, performs filtering processing in the spatial domain, and uses a non-linear filter method to perform motion estimation difference images before DCT.
  • the filtering is performed twice.
  • a threshold filter can be used to reduce or eliminate random noise
  • a cross-shaped median filter can be used to eliminate high-frequency impulse noise, which can reduce the total number of code words by 10% -20%.
  • the existing denoising methods generally need to add a 1-D or 2-D denoising filtering process to the original processing flow, and the filtering calculation generally takes more time, thereby increasing the calculation burden.
  • a denoising method that requires less time is calculated, and the image content must be protected as much as possible.
  • the cabinet filtering method in the prior art is a method for pre-denoising the image with a small amount of calculation.
  • the basic principle is that in the quantization of the DCT coefficients of the image block of the CIF format image at the transmitting end, each DCT coefficient is processed in a certain order If the DCT coefficient value of this process is less than or equal to a preset threshold, the DCT coefficient value is set to 0. If the DCT coefficient value of this process is greater than the preset threshold, the DCT coefficient value is maintained. The original value is unchanged.
  • This image pre-denoising method is essentially a fixed threshold filtering method. Its disadvantages are: the choice of threshold is very contradictory. If the threshold is selected smaller, higher image quality can be achieved, but it will be due to DCT. Too little coefficient 0 results in unsatisfactory image compression efficiency; if the threshold is chosen to be large, satisfactory image compression efficiency can be achieved, but it will not reach high due to too many values of DCT coefficients Image quality. Summary of the invention
  • An object of the present invention is to provide a method for image pre-denoising, which essentially incorporates an adaptive threshold filtering method. Compared with the fixed threshold filtering method in the prior art, the method can not reduce image quality. To further improve image compression efficiency.
  • An image pre-denoising method which is characterized by including at least: in the DCT coefficient quantization of the image block CIF format image of the transmitting end, each DCT coefficient is processed in a certain order in sequence.
  • the processed DCT coefficient value is less than or equal to a threshold corresponding to the current processing, then the DCT coefficient value is set to 0, and the threshold is increased to be used as the threshold corresponding to the next DCT coefficient processing. If the threshold is greater than a preset upper limit, it will be equal to the upper limit; if the DCT coefficient value of this process is greater than the threshold corresponding to this process, the threshold will be restored to a preset initial threshold It is used as the threshold corresponding to the next DCT coefficient processing.
  • the image pre-denoising method of the present invention further includes: sending a video image from CCIR601 (CCIR is an abbreviation of International Radiocommunication Advisory Committee, CCIR601 is a 601 standard proposed by CCIR at the transmitting end, and defines corresponding to 525 lines and 625
  • CCIR601 is an abbreviation of International Radiocommunication Advisory Committee
  • CCIR601 is a 601 standard proposed by CCIR at the transmitting end, and defines corresponding to 525 lines and 625
  • a low-pass FIR filter with a cutoff frequency of less than 0.5 ⁇ is used to filter the luminance component and chrominance component in the horizontal and vertical directions, respectively.
  • the low-pass FIR filter is a one-dimensional low-pass FIR filter.
  • the cut-off frequency of the low-pass FIR filter may be between 0.25 ⁇ and 0,4 ⁇ .
  • the image pre-denoising method of the present invention further includes: using a low-pass FIR filter with a cutoff frequency of less than 0.5 ⁇ , respectively dividing the luminance component and the chrominance in the horizontal and vertical directions.
  • a two-dimensional median filter is further used to filter the image.
  • the DCT coefficient value of the current process is less than or equal to the threshold corresponding to the current process
  • the DCT coefficient value is made 0, and the threshold is added to 1 as the DCT coefficient corresponding to the next time. Processing threshold.
  • the sequence of processing each DCT coefficient in sequence is a zigzag starting from the upper left corner of the DCT coefficient table.
  • the initial threshold and the upper threshold may be set in advance as follows:
  • the initial threshold is equal to kQ
  • the upper threshold is equal to 1.5kQ
  • Q is a quantization level
  • k is determined according to a channel bandwidth.
  • a constant in the range of 0 to 1.
  • the method for image pre-denoising provided by the present invention can be understood as a method combining an adaptive threshold filtering, and compared with the fixed threshold filtering method in the prior art, the method can be performed without reducing image quality. Filter out noise more effectively, thereby further improving image compression coding efficiency.
  • the above-mentioned non-linear noise value filtering processing part of the edge information of the image will be lost while the noise is filtered, but according to the human eye, the sensitivity to the low-frequency component in the image is much higher than the visual characteristics of the high-frequency component, as long as Within the permissible range of human visual perception, the above-mentioned adaptive threshold processing can be used to obtain a better image shield amount.
  • Another advantage of adaptive threshold filtering is that the amount of calculation is relatively small, which is very suitable for video service systems with high real-time and interactive requirements.
  • Figure 1 shows the distribution of CT coefficients for 8 x 8 image blocks in H.26X (i.e. H.261, H.263, etc.) format;
  • FIG. 2 shows a basic architecture of an image pre-denoising method according to a preferred embodiment of the present invention
  • FIG. 3 shows a processing flow of the adaptive threshold filtering in the preferred embodiment of FIG. 2.
  • FIG. 4 shows a basic architecture of an image pre-noising method according to another preferred embodiment of the present invention.
  • Fig. 2 shows the basic architecture of an image pre-noise method according to a preferred embodiment of the present invention, that is, the position of the filtering process in the H.26X video compression process.
  • this embodiment uses two filters: a one-dimensional low-pass FIR filter and an adaptive noise filter.
  • the one-dimensional low-pass FIR filter is used as a linear high-frequency impulse noise filter.
  • the adaptive threshold filter is a non-linear random noise filter, and performs filtering processing in the space domain and the frequency domain, respectively.
  • the one-dimensional low-pass FIR filter and the adaptive threshold filter in this embodiment will be described separately.
  • a one-dimensional low-pass FIR (Finite Impulse Response) filter is used as a high-frequency impulse noise filter, and the luminance components are horizontally and vertically respectively.
  • the chroma components are filtered, and the high-frequency pulse noise included in the image is removed while the image format is being converted.
  • the corresponding high-frequency part in the image will also be filtered out, but according to the human eye's sensitivity to the low-frequency component in the image is much higher than the visual characteristics of the high-frequency component, as long as it is within the allowable range of human visual perception, an appropriate A low-pass filter with a cut-off frequency can obtain better image quality.
  • the image can be regarded as a one-dimensional signal for spectrum analysis, and by comparing the fitting curves of the spectrum of the image before the H.26X compression and the reconstructed image, It can be seen that there is a large difference between the two, and the fitted curve of the spectrum of the image filtered by the filter with a small cut-off frequency ( ⁇ 0.5 ⁇ ) and the reconstructed image However, the difference is very small. According to the former, the difference can help us determine the range of the cutoff frequency. Generally, it can be taken as 0.25 ⁇ to 0.4 ⁇ .
  • the number of codewords of the image after H.26X compression can be reduced by 10% -20%.
  • the image quality is basically the same at the same quantization level, but in actual H.26X In the system, the general control strategy will reduce the quantization level when the buffer occupancy is reduced, thereby improving the image quality.
  • Figure 1 shows the DCT coefficient distribution of 8 x 8 image blocks in H.26X format.
  • the DCT transform discrete cosine transform
  • the low-frequency component is located in the upper left and the high-frequency component is located in the lower right.
  • the frequency component corresponds to the detailed content of the image
  • the high frequency component corresponds to the edge information of the image and the noise interspersed in the image, such as random noise and high frequency impulse noise. Since most of the high frequency impulse noise has been filtered out after the high frequency noise filtering, the random noise is mainly filtered here.
  • the adaptive threshold filter in this embodiment sequentially processes 64 DCT coefficients in accordance with a zig-zag scanning order.
  • the detailed processing flow is shown in Figure 3.
  • * DCT_Coeff is a pointer to the DCT coefficient
  • the threshold Thresh of the DCT coefficient is related to the quantization level 0 value.
  • Thresh-Max is the upper limit of the set threshold.
  • Thresh-Max is set to 1.5kQ, and the initial cabinet value is set to kQ.
  • the value of k ranges from 0.0 to 1.0.
  • the value is determined according to the channel bandwidth.
  • the channel bandwidth is small.
  • the coefficient value is less than or equal to the threshold Thresh, then the coefficient value is taken as 0, and the threshold value is increased by 1, so that the processing is conducive to the occurrence of zero contiguous situations as much as possible, thereby reducing the number of encoding code words and improving compression efficiency.
  • threshold Thresh When the value is greater than the upper threshold Thresh_Max, the value is taken as Thresh_Max. If the coefficient value is greater than the threshold Thresh, the threshold Thresh is restored to the initial threshold kQ, which is beneficial for protecting important edge information in the image.
  • the above-mentioned non-linear noise value filtering processing part of the edge information of the image will be lost while the noise is filtered, but according to the human eye, the sensitivity to the low-frequency component in the image is much higher than the visual characteristics of the high-frequency component, as long as Within the allowable range of human visual perception, the above-mentioned adaptive threshold processing can be used to obtain better image quality.
  • This embodiment is an efficient denoising method combining linear filtering and nonlinear filtering, combining spatial domain filtering and frequency domain filtering to remove video image noise. It uses a linear filter and a non-linear filter respectively, which can reduce or eliminate the high-frequency impulse noise and random noise existing in the video image at a small computational cost. It has the advantages of convenient use and low calculation burden. For the first denoising, only a filter with a lower cut-off frequency is selected for format conversion, and it does not increase any calculation burden. The second denoising uses an adaptive domain value filter during DCT quantization. As can be seen from Fig. 3, the amount of calculation required is very small. The denoising method in this embodiment can reduce the number of codewords by 10% to 30%, thereby improving video compression efficiency and image quality.
  • Fig. 4 shows a basic architecture of an image pre-noising method according to another preferred embodiment of the present invention.
  • a 3 ⁇ 3 two-dimensional median filter is added on the basis of the previous embodiment, and it is set before conversion to a CIF format image.
  • the median filter should be implemented in hardware or DSP software. Because the median filter has the characteristics of better removing impulse noise and maintaining the edge of the image, combined with the subsequent adaptive threshold filtering processing, a good image denoising effect can be obtained.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Computing Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)
  • Color Television Systems (AREA)
  • Image Processing (AREA)
  • Facsimile Image Signal Circuits (AREA)

Description

图像预去噪的方法
技术领域
本发明涉及多媒体技术中的图像通信领域， 特别是涉及发送端图像 预去噪的方法。 发明背景
视频图像包含的数据量一般很大，不便于直接进行视频图像通讯（如 会议电视、 可视电话、 远程教学等）与存储（如多媒体数据库、 VCD、 DVD等），特别是对视讯业务系统来说，除了视频外，还包括音频、 T.120 中的数据、 控制信息等内容， 并且对实时性、 交互性和图像盾量要求很 高， 故一般要对视频图像数据进行较大比率的压缩， 以便获得满意的视 频效果。 现有的视频图像压缩国际标准有国际电信联盟 ITU-T制定的 H.261、 H.263和国际标准化组织 ISO制定的 MPEG1、 MPEG2、 MPEG4, 它们分别包含了当今的一些重要压缩技术，其中有许多共同的地方，如： 都采用 CIF (公共中间格式）格式、 都采用运动补偿和 DCT (离散余弦 变換） 混合模式等。
在根据例如以上的这些标准实现具体的视频系统时， 由于所获取的 视频图像中一般都不可避免地包含了一些不需要的噪声， 如高频脉冲噪 声 （图像小区域内亮度分量和色度分量的突然大幅度变化值）和随机噪 声 （如模 /数转换所引起的噪声、 信号采样时的量化噪声）， 如果不在压 缩编码前进行去噪处理， 图像的压缩效率将会受到很大影响。 因此， 为 了得到较好的压缩效率和图像质量， 一般都要进行必要的前处理和后处 逑， 以去掉或减小夹杂在图像中的噪声。
已有的去噪方法很多， 一般都是采用适当的滤波方法， 可在空间域 或频率域进行， 按是否线性可分为： 线性滤波， 如 1-D FIR (—维有限脉 冲响应）滤波、 2-D FIR (二维有限脉冲响应）滤波等； 非线性滤波， 如 中值滤波、 阔值滤波等。
参考文件 [1]， 即 US5787203: "Method and system for filtering compressed video images" ， 中公开的方法， 在空间域内进行滤波处理， 采用非线性滤波器方法在 DCT前对运动估计后的差值图像进行两次滤 波， 首先通过阈值滤波器可减小或消除随机噪声， 然后通过十字形中值 滤波器可消除高频脉冲噪声， 可以使总码字数减小 10%-20%。 参考文件 [2], ^PUS5325125: "Intra-frame filter for video compression systems" , 中公开的方法，在压缩前直接对图像帧本身进行线性滤波处理，采用 2-D 滤波器消除图像对角线方向的高频分量和高频脉沖噪声， 从而减小编码 码字数， 提高压缩效率。
为有效地去掉噪声， 现有的去噪方法一般要在原处理流程中增加一 个 1-D或 2-D去噪滤波处理过程， 而滤波计算一般耗时较多， 从而增加了 计算负担， 对于实时性、 交互性要求很高的会议电视系统而言， 需要计 算耗时较少的去噪方法， 并且要尽量地保护图像内容。 现有技术中的阁 值滤波法就是一种计算量较小的图像预去噪的方法， 其基本原理是在发 送端 CIF格式图像的图像块 DCT系数量化中， 依照一定顺序依次处理各 个 DCT系数， 如果本次处理的 DCT系数值小于或等于一个预先设定的阈 值， 则令该 DCT系数值为 0， 如果本次处理的 DCT系数值大于该预先设 定的阈值， 则该 DCT系数值保持原值不变。 这种图像预去噪方法本质上 是一种固定阔值滤波的方法， 其缺点在于： 阔值的选取非常的矛盾， 如 果阈值选的较小则可以达到较高的图像质量， 但会由于 DCT系数的 0值 过少而导致达不到满意的图像压缩效率； 如果阈值选的较大则可以达到 满意的图像压缩效率， 但会由于 DCT系数的 0值过多而导致达不到较高 的图像质量。 发明内容
本发明的目的在于提供一种图像预去噪的方法， 其本质上结合了一 种自适应阈值滤波的方法， 相比较于现有技术中的固定阔值滤波的方 法， 可以在不降低图像质量的前提下进一步提高图像压缩效率。
本发明的目的是这样实现的： 一种图像预去噪的方法， 其特征在于 至少包含： 在发送端 CIF格式图像的图像块 DCT系数量化中， 依照一定 顺序依次处理各个 DCT系数， 如果本次处理的 DCT系数值小于或等于一 个对应于本次处理的阔值， 则令该 DCT系数值为 0, 且令该阈值增大后 作为对应于下一次 DCT系数处理的阈值， 如果增大后的阈值大于预先设 定的一个闹值上限则令其等于该阔值上限； 如果本次处理的 DCT系数值 大于对应于本次处理的阔值， 则令该阈值恢复为预先设定的一个初始阈 值后作为对应于下一次 DCT系数处理的阈值。
较佳地， 本发明的图像预去噪方法进一步包含： 在发送端将视频图 像从 CCIR601 ( CCIR是国际无线电通信咨询委员会的缩写， CCIR601是 CCIR提出的 601标准， 定义了对应于 525行和 625行电视系统的电视演播 的数字视频标准）格式转换为 CIF格式时， 采用截止频率小于 0.5 π的低 通 FIR滤波器， 分别在水平和垂直方向对亮度分量、 色度分量进行滤波 处理。
较佳地， 所述的低通 FIR滤波器是一维低通 FIR滤波器。
较佳地， 所述的低通 FIR滤波器的截止频率可以在 0.25 π至 0,4 π之 间。
较佳地， 本发明的图像预去噪方法进一步包含： 在采用截止频率小 于 0.5 π的低通 FIR滤波器， 分别在水平和垂直方向对亮度分量、 色度分 量进行滤波处理后， 进一步采用一个二维中值滤波器对图像进行滤波处 理。
较佳地， 在本次处理的 DCT系数值小于或等于对应于本次处理的阈 值的情况下， 则令该 DCT系数值为 0, 且令该阔值加 1后作为对应于下一 次 DCT系数处理的阈值。
较佳地， 所述依次处理各个 DCT系数的顺序为从 DCT系数表的左上 角开始的之字型。
较佳地， 所述的初始阔值和所述的阈值上限可以这样预先设定： 初 始阔值等于 kQ, 阔值上限等于 1.5kQ, 其中 Q为量化级， k为根据信道带 宽的大小确定的取值范围在 0至 1之间的一个常数。
本发明提供的图像预去噪的方法， 可以理解为是结合了一种自适应 阈值滤波的方法， 相较于现有技术中的固定阔值滤波的方法， 可以在不 降低图像质量的前提下更有效地滤除噪声， 从而进一步提高图像压缩编 码效率。 当然， 进行上述非线性闹值滤波处理后， 在滤除噪声的同时图 像的部分边缘信息也将丟失， 但根据人眼对图像中低频分量的敏感性大 大高于高频分量的视觉特性， 只要在人的视觉感知允许范围内， 采用上 述自适应阈值处理， 可以获得较好的图像盾量。 自适应阈值滤波的另一 个优点是计算量相对较小， 非常适合对于实时性、 交互性要求很高的视 讯业务系统。 附图简要说明
图 1示出了 H.26X (即 H.261、 H.263等）格式的 8 x 8图像块的 £)CT系 数分布；
图 2示出了依照本发明的一个较佳实施例的图像预去噪方法的基本 架构； 图 3示出了图 2的较佳实施例中的自适应阈值滤波的处理流程。 图 4示出了依照本发明的另一个较佳实施例的图像预去噪方法的基 本架构。 实施本发明的方式
下面结合附图对本发明进行详细描述。 所应理解的是， 其仅用于对 本发明的阐述说明而非限制。
图 2示出了依照本发明的一个较佳实施例的图像预去噪方法的基本 架构，也即滤波处理在 H.26X视频压缩过程中所处的位置。如图 2所示，本 实施例采用了两个滤波器： 一维低通 FIR滤波器和自适应闹值滤波器， 其中一维低通 FIR滤波器作为线性的高频脉冲噪声滤波器， 自适应阔值 滤波器作为非线性的随机噪声滤波器， 分别在空间域和频率域进行滤波 处理。 以下将对本实施例中的一维低通 FIR滤波器和自适应阔值滤波器 分别进行说明。
如图 2所示， 在将视频图像从 CCIR 601格式转换为 CIF格式时， 采用 一维低通 FIR (有限脉冲响应） 滤波器作为高频脉冲噪声滤波器， 分别 在水平和垂直方向对亮度分量、 色度分量进行滤波处理， 在进行图像格 式转换的同时， 将图像中所包含的高频脉冲噪声去掉。 当然， 图像中的 相应高频部分也将被滤掉， 但是， 根据人眼对图像中低频分量的敏感性 大大高于高频分量的视觉特性， 只要在人的视觉感知允许范围内， 采用 适当截止频率的低通滤波器， 可以获得较好的图像质量。
针对本实施例中一维 FIR低通滤波器的截止频率的选取， 可以将图 像看成一维信号进行频谱分析， 通过对 H.26X压缩前图像和重构图像的 频谱的拟合曲线进行对比， 可以看出两者差别较大， 而经过较小截止频 率 （<0.5 π ) 的滤波器滤波后的图像和重构图像的频谱的拟合曲线之间 却相差甚小， 根据前者差值情况可以帮助我们确定截止频率的范围， 一 般可取为 0.25 π至 0.4 π。 以七阶滤波器为例， 本实施例中的低通 FIR滤 波器可以是截止频率为 0.4 π的滤波器 h={-2/256, 12/256, 66/256, 104/256, 66/256， 12/256， -2/256}。
经一维 4氏通 FIR滤波后， 图像经过 H.26X压缩后的码字数可减小 10%-20%, 如在相同量化级时图像质量基本上没有什么差别， 但在实际 的 H.26X系统中， 一般的控制策略都会在緩沖区占用量减小时降低量化 级， 从而可以提高图像质量。
以下针对本实施例中的自适应阔值滤波进行说明。 图 1示出了 H.26X 格式的 8 x 8图像块的 DCT系数分布。 如图 1所示， 8x8图像块差值信号经 过 DCT变换（离散余弦变换）后， 低频分量位于左上部， 高频分量则位 于右下部。 4氐频分量对应于图像的细节内容， 而高频分量则对应于图像 的边缘信息和夹杂于图像中的噪声， 例如随机噪声和高频脉冲噪声。 由 于经过高频^ c冲噪声滤波后已经滤除了大部分高频脉冲噪声， 因此， 在 这里主要是要滤除随机噪声。
本实施例中的自适应阈值滤波器，按照之字型（ Zig-zag )扫描顺序， 依次处理 64个 DCT系数。其详细处理流程如图 3所示，其中， *DCT一 Coeff 为指向 DCT系数的指针， DCT系数的阈值 Thresh与量化级0值相关， 与 固定阈值相比， 更适应于图像内容的变化和有利于图像重要内容的保 护。 Thresh一 Max为设定的阈值上限， Thresh一 Max取为 1.5kQ, 初始阁值 取为 kQ， 其中 k值的取值范围为 0.0至 1.0, 其值根据信道带宽的大小来确 定， 信道带宽小时取较大的 k值， 例如低于 384kbps时可取为 1.0, 信道带 宽大时取较小的 k值， 例如 2Mbps时可取为 0.4。 如果系数值小于或等于 阈值 Thresh, 则将系数值取为 0， 并将阔值加 1, 这样处理有利于尽量多 地出现连零的情况，从而减小编码码字数，提高压缩效率。 当阈值 Thresh 增加到大于阈值上限 Thresh一 Max时， 则取为 Thresh— Max值。 如果系数值 大于阈值 Thresh, 则将阔值 Thresh恢复为初始阈值 kQ， 这样处理有利于 保护图像中重要的边缘信息。
当然， 进行上述非线性闹值滤波处理后， 在滤除噪声的同时图像的 部分边缘信息也将丢失， 但根据人眼对图像中低频分量的敏感性大大高 于高频分量的视觉特性， 只要在人的视觉感知允许范围内， 采用上述自 适应阔值处理， 可以获得较好的图像质量。
本实施例是一种线性滤波和非线性滤波相结合、 空间域滤波和频率 域滤波相结合、 以去掉视频图像噪声的高效的去噪方法。 其分别采用一 个线性滤波器和一个非线性滤波器， 可以以较小的计算代价减小或消除 视频图像中存在的高频脉沖噪声和随机噪声。 具有使用方便、 计算负担 小等优点。 第一次去噪只是选用了一个截止频率较低的滤波器以进行格 式转换， 并不增加任何计算负担。 第二次去噪在 DCT量化时使用自适应 域值滤波器， 由图 3可知， 需要增加的计算量非常小。 本实施例的去噪 方法可以使码字数减小 10%到 30%，从而提高视频压缩效率和图像质量。
图 4示出了依照本发明的另一个较佳实施例的图像预去噪方法的基 本架构。 该较佳实施例是在前一实施例的基础上再增加一个 3X3二维中 值滤波器， 设置在转换为 CIF格式图像之前。 考虑到运算量较大， 该中 值滤波器宜采用硬件实现方式或 DSP软件实现方式。 由于中值滤波器具 有较好地去掉脉冲噪声和保持图像边缘的特点， 再结合后面的自适应阈 值滤波处理， 这样就可以得到艮好的图像去噪效果。
以上实施例仅用以说明本发明而非限制， 尽管参照较佳实施例对本 发明进行了详细说明， 本领域的普通技术人员应当理解， 可以对本发明 进行修改或者等同替换， 而不脱离本发明的精神和范围， 其均应涵盖在 本发明的权利要求范围当中。

Claims

权利要求书
1、 一种图像预去噪的方法，其特征在于至少包含： 在发送端 CIF 格式图像的图像块 DCT系数量化中， 依照一定顺序依次处理各个 DCT 系数： 如果本次处理的 DCT 系数值小于或等于一个对应于本次处理的 阈值， 则令该 DCT 系数值为 0, 且令该阔值增大后作为对应于下一次 DCT系数处理的阈值，如果增大后的阈值大于预先设定的一个阈值上限 则令其等于该阈值上限； 如果本次处理的 DCT 系数值大于对应于本次 处理的阈值， 则令该闹值恢复为预先设定的一个初始阈值后作为对应于 下一次 DCT系数处理的阈值。
2、 如权利要求 1所述的图像预去噪的方法， 其特征在于进一步 包含：在发送端将视频图像从 CCIR601格式转换为 CIF格式时，采用截 止频率小于 0.5 π的低通 FIR滤波器， 分别在水平和垂直方向对亮度分 量、 色度分量进行滤波处理。
3、 如权利要求 2所述的图像预去噪的方法， 其特征在于， 所述 的低通 FIR滤波器是一维低通 FIR滤波器。
4、 如权利要求 2所述的图像预去噪的方法， 其特征在于， 所述 的低通 FIR滤波器的截止频率可以在 0.25 π至 0.4 π之间。
5、 如权利要求 2所述的图像预去噪的方法， 其特征在于进一步 包含： 在采用截止频率小于 0.5 π的低通 FIR滤波器， 分别在水平和垂 直方向对亮度分量、 色度分量进行滤波处理后， 进一步采用一个二维中 值滤波器对图像进行滤波处理。
6、 如权利要求 1所述的图像预去噪的方法， 其特征在于， 在本 次处理的 DCT 系数值小于或等于对应于本次处理的阔值的情况下， 则 令该 DCT系数值为 0, 且令该阔值加 1后作为对应于下一次 DCT系数 处理的阈值。
7、 如权利要求 1所述的图像预去噪的方法， 其特征在于， 所述 依次处理各个 DCT 系数的顺序为从 DCT 系数表的左上角开始的之字 型。
8、 如权利要求 1所述的图像预去噪的方法， 其特征在于， 所述 的初始阔值和所述的阔值上限可以这样预先设定： 初始阈值等于 kQ， 阔值上限等于 1.5kQ， 其中 Q为量化级， k为根据信道带宽的大小确定 的取值范围在 0至 1之间的一个常数。
PCT/CN2002/000110 2001-02-27 2002-02-25 Procede de presuppression du bruit d'une image Ceased WO2002078350A1 (fr)

Priority Applications (3)

Application Number Priority Date Filing Date Title
DE60220380T DE60220380T2 (de) 2001-02-27 2002-02-25 Verfahren zum vorlöschen von rauschen eines bildes
EP02703466A EP1365590B1 (en) 2001-02-27 2002-02-25 A method for pre-deleting noise of image
US10/601,732 US7194030B2 (en) 2001-02-27 2003-06-24 Method for pre-suppressing noise of image

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN01109026.X 2001-02-27
CNB01109026XA CN1167034C (zh) 2001-02-27 2001-02-27 图像预去噪的方法

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US10/601,732 Continuation US7194030B2 (en) 2001-02-27 2003-06-24 Method for pre-suppressing noise of image

Publications (1)

Publication Number Publication Date
WO2002078350A1 true WO2002078350A1 (fr) 2002-10-03

Family

ID=4657655

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2002/000110 Ceased WO2002078350A1 (fr) 2001-02-27 2002-02-25 Procede de presuppression du bruit d'une image

Country Status (7)

Country Link
US (1) US7194030B2 (zh)
EP (2) EP1821548B1 (zh)
CN (1) CN1167034C (zh)
DE (2) DE60220380T2 (zh)
HK (1) HK1047647B (zh)
RU (1) RU2260922C2 (zh)
WO (1) WO2002078350A1 (zh)

Families Citing this family (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7869500B2 (en) * 2004-04-27 2011-01-11 Broadcom Corporation Video encoder and method for detecting and encoding noise
US7949051B2 (en) * 2004-05-26 2011-05-24 Broadcom Corporation Mosquito noise detection and reduction
JP4359773B2 (ja) * 2004-06-22 2009-11-04 ソニー株式会社 画像圧縮処理装置、画像圧縮処理方法および画像圧縮処理プログラム
US7590300B2 (en) * 2004-06-24 2009-09-15 Intel Corporation Image filtering methods and apparatus
US8031768B2 (en) * 2004-12-15 2011-10-04 Maxim Integrated Products, Inc. System and method for performing optimized quantization via quantization re-scaling
US7471844B2 (en) * 2004-12-27 2008-12-30 Intel Corporation Method, apparatus and system for multi-feature programmable tap filter image processing
US8218634B2 (en) * 2005-01-13 2012-07-10 Ntt Docomo, Inc. Nonlinear, in-the-loop, denoising filter for quantization noise removal for hybrid video compression
CN101257630B (zh) * 2008-03-25 2012-06-06 浙江大学 结合三维滤波的视频编码方法和装置
US8326075B2 (en) 2008-09-11 2012-12-04 Google Inc. System and method for video encoding using adaptive loop filter
KR20110062008A (ko) * 2009-12-02 2011-06-10 삼성전자주식회사 화상형성장치 및 그 화상 노이즈 처리 방법
CN102740065B (zh) * 2011-03-31 2015-11-25 富士通株式会社 图像编码方法和系统
US8885706B2 (en) 2011-09-16 2014-11-11 Google Inc. Apparatus and methodology for a video codec system with noise reduction capability
US9131073B1 (en) 2012-03-02 2015-09-08 Google Inc. Motion estimation aided noise reduction
US9344729B1 (en) 2012-07-11 2016-05-17 Google Inc. Selective prediction signal filtering
US10102613B2 (en) 2014-09-25 2018-10-16 Google Llc Frequency-domain denoising
CN107895351B (zh) * 2017-10-30 2019-08-20 维沃移动通信有限公司 一种图像去噪方法及移动终端
CN109640166A (zh) * 2018-08-13 2019-04-16 张利军 基于动感程度的片尾曲选择方法
CN110295345A (zh) * 2019-07-30 2019-10-01 浦江和平真空镀膜有限公司 一种真空镀膜机
CN113645468A (zh) * 2021-08-19 2021-11-12 广东博华超高清创新中心有限公司 动态视觉传感器滤波加速控制方法、系统、设备及应用
CN114358040A (zh) * 2021-11-29 2022-04-15 国网安徽省电力有限公司马鞍山供电公司 一种基于自适应估算阈值法的扰动信号去噪方法

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2000042772A1 (en) * 1999-01-15 2000-07-20 Koninklijke Philips Electronics N.V. Coding and noise filtering an image sequence

Family Cites Families (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
SU965016A1 (ru) * 1981-03-12 1982-10-07 Минское Производственное Объединение "Горизонт" Устройство фильтрации телевизионного сигнала
SU1438023A1 (ru) * 1987-04-21 1988-11-15 Грузинский политехнический институт им.В.И.Ленина Двумерный адаптивный восстанавливающий фильтр
JPH0310480A (ja) * 1989-06-07 1991-01-18 Toshiba Corp テレビジョン信号伝送装置およびテレビジョン信号受信装置
US5121216A (en) * 1989-07-19 1992-06-09 Bell Communications Research Adaptive transform coding of still images
US5241395A (en) * 1989-08-07 1993-08-31 Bell Communications Research, Inc. Adaptive transform coding using variable block size
US5325125A (en) 1992-09-24 1994-06-28 Matsushita Electric Corporation Of America Intra-frame filter for video compression systems
JP2894137B2 (ja) * 1993-01-26 1999-05-24 ケイディディ株式会社 動画像符号化におけるプリフィルタ制御方法および装置
JP3466705B2 (ja) * 1993-05-28 2003-11-17 ゼロックス・コーポレーション 圧縮画像の圧縮解除方法
JPH07107462A (ja) * 1993-09-29 1995-04-21 Oki Electric Ind Co Ltd 動画像符号化方法
KR960039992A (ko) * 1995-04-29 1996-11-25 배순훈 블록화 현상 제거 장치 및 이를 이용한 동영상 복호화기
JP3353604B2 (ja) * 1995-08-09 2002-12-03 ソニー株式会社 動画像符号化方法及び装置、並びに信号記録媒体
US6064450A (en) * 1995-12-06 2000-05-16 Thomson Licensing S.A. Digital video preprocessor horizontal and vertical filters
US5787203A (en) 1996-01-19 1998-07-28 Microsoft Corporation Method and system for filtering compressed video images
US5832115A (en) * 1997-01-02 1998-11-03 Lucent Technologies Inc. Ternary image templates for improved semantic compression
CA2316950C (en) * 1998-11-04 2004-06-08 Mitsubishi Denki Kabushiki Kaisha Image encoder and image decoder limiting encoded data with respect to frequency domains
US6370502B1 (en) * 1999-05-27 2002-04-09 America Online, Inc. Method and system for reduction of quantization-induced block-discontinuities and general purpose audio codec
US7206453B2 (en) * 2001-05-03 2007-04-17 Microsoft Corporation Dynamic filtering for lossy compression
US7003174B2 (en) * 2001-07-02 2006-02-21 Corel Corporation Removal of block encoding artifacts
JP3972751B2 (ja) * 2001-08-27 2007-09-05 コニカミノルタビジネステクノロジーズ株式会社 画像処理装置
US7054367B2 (en) * 2001-12-31 2006-05-30 Emc Corporation Edge detection based on variable-length codes of block coded video

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2000042772A1 (en) * 1999-01-15 2000-07-20 Koninklijke Philips Electronics N.V. Coding and noise filtering an image sequence

Also Published As

Publication number Publication date
RU2003126189A (ru) 2005-03-10
HK1047647B (zh) 2005-05-20
EP1821548A2 (en) 2007-08-22
EP1821548A3 (en) 2008-03-12
US7194030B2 (en) 2007-03-20
US20040062316A1 (en) 2004-04-01
HK1047647A1 (zh) 2003-02-28
DE60220380T2 (de) 2008-01-24
CN1167034C (zh) 2004-09-15
EP1365590B1 (en) 2007-05-30
CN1372224A (zh) 2002-10-02
DE60235211D1 (de) 2010-03-11
DE60220380D1 (de) 2007-07-12
EP1365590A1 (en) 2003-11-26
RU2260922C2 (ru) 2005-09-20
EP1365590A4 (en) 2004-05-12
EP1821548B1 (en) 2010-01-20

Similar Documents

Publication Publication Date Title
WO2002078350A1 (fr) Procede de presuppression du bruit d&#39;une image
KR100863166B1 (ko) 블록 기반 디지털 비디오의 역블럭화 및 에지 강화를 위한 방법, 장치, 전자 디바이스, 컴퓨터 판독가능 매체, 및 프로세서
US5757969A (en) Method for removing a blocking effect for use in a video signal decoding apparatus
JPH06225284A (ja) ビデオ圧縮システムのためのフレーム内フィルタ及びフィルタリング方法
JP2010004555A (ja) 画像復号方法及び画像復号装置
WO2007033551A1 (fr) Procede de filtrage a boucle destine a une operation de codage d&#39;image
KR20050074906A (ko) 루프 필터링 방법 및 루프 필터
US7822125B2 (en) Method for chroma deblocking
JP5362723B2 (ja) 圧縮画像ノイズ除去装置と再生装置
JP3800435B2 (ja) 動画像信号処理装置
JP2007006390A (ja) 画像処理装置、画像処理方法及びプログラム
JPH09149417A (ja) 動画像信号復号化装置
JP3169263B2 (ja) 動画像データの符号化装置
JPH0865667A (ja) 画像信号符号化方法
KR100764767B1 (ko) 부호화 모드 선택 방법
Cahill et al. Locally adaptive deblocking filter for low bit rate video
JPH06217296A (ja) 適応的イントラ／インタモード圧縮を用いた映像信号符号化装置
JP2005260989A (ja) 画像処理装置及び画像処理方法
Zheng et al. Overview and performance analysis of AVS mobility features
Goto et al. A Deblocking Method using Wavelet Transform for MPEG-II format
JPH07274165A (ja) 画像情報圧縮装置
JPH04297192A (ja) 動画像の適応フレーム間直交変換符号化方法
JPH0686254A (ja) 動画像データの高能率符号化装置
JP2002238053A (ja) 画像エンファシス符号化装置及び画像エンファシス符号化方法
HK1070223B (zh) 基於边界强度的自适应过滤

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY BZ CA CH CO CR CU CZ DE DK DM DZ EC EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NO NZ OM PH PL PT RO RU SD SE SG SI SK SL TJ TM TN TR TT TZ UA UG US UZ VN YU ZA ZM ZW

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
WWE Wipo information: entry into national phase

Ref document number: 2002703466

Country of ref document: EP

WWE Wipo information: entry into national phase

Ref document number: 10601732

Country of ref document: US

WWP Wipo information: published in national office

Ref document number: 2002703466

Country of ref document: EP

REG Reference to national code

Ref country code: DE

Ref legal event code: 8642

NENP Non-entry into the national phase

Ref country code: JP

WWW Wipo information: withdrawn in national office

Country of ref document: JP

WWG Wipo information: grant in national office

Ref document number: 2002703466

Country of ref document: EP