WO2013004161A1 - Method and device for classifying pixel of video image - Google Patents

Abstract

Disclosed are a method and device for classifying a pixel of a video image. An embodiment of the method comprises: acquiring a directional value corresponding to any pixel or pixel block in the video image; acquiring a complexity value corresponding to the pixel or pixel block; where the classification number of the directional value is greater than or equal to the classification number of the complexity value; and acquiring a classification result of the pixel on the basis of the directional value and of the complexity value of the pixel. Embodiments of the present invention increase the proportion of a directional characteristic classification in the process of classifying the pixel or pixel block of adaptive filtering, thus improving the rationality of a filter design, and allowing for improved coding efficiency.

Description

 Method and device for pixel classification of video images

 The present application claims priority to Chinese Patent Application No. 201110184115.7, entitled "Pixel Classification Method and Apparatus for Video Image", filed on July 1, 2011, the entire contents of which are incorporated by reference. In this application. Technical field

 The present invention relates to the field of digital signal processing technologies, and more particularly to a pixel classification method and apparatus for video images. Background technique

 In the field of digital communications, voice, image, audio, and video transmissions have a wide range of application requirements, such as mobile phone calls, audio and video conferencing, broadcast television, multimedia entertainment, and the like. With the development of network technology, video on demand, Internet TV, videophone, etc. have become the main business of broadband networks, and these services will become the main business of the 3rd generation (3G, the 3rd Generation) wireless network. In order to reduce the resources occupied during video signal storage or transmission, the video signal is compressed at the transmitting end and transmitted to the receiving end, and the receiving end recovers the video signal and plays it by decompressing. The sender or receiver here can be a mobile phone, a digital phone terminal, a wireless device, a personal data assistant (PDA), a handheld or portable computer, a GPS receiver/navigator, a camera, an audio/video player, a video camera, a video recorder. , monitoring equipment, etc.

ALF (Adaptive Loop Filter) technology is to train multiple reconstructed images and original input images to train multiple sets of filter coefficients, filter the reconstructed images, improve the reconstructed image quality, and simultaneously encode loops. The internal performance is improved by improving motion compensation prediction performance. In the latest coding standard HEVC, ALF is applied to the reconstructed image deblocking filtered output image. First, HEVC defines a pixel classification method based on the local statistical characteristics of pixels. This classification method divides all the pixels in the image into 16 categories, and uses the pixels in each category to train the ALF coefficients for the pixels. Then, according to the coded rate distortion optimal criterion, the ALF coefficient and the pixel class are combined, the filter coefficients are retrained for the merged class, and finally the number of filters and corresponding coefficients that need to be written into the code stream are obtained at the encoding end, And use these filter coefficients to filter the image. In order to obtain better filtering performance, three kinds of diamond filter structures are designed in HEVC, as shown in Figure 1. 5x5, 7x7 and 9x7.

 In order to better remove the compression noise, the ALF coefficient can utilize the statistical characteristics of the local direction and variance of the image in the existing HEVC, calculate the direction value and the variance value of the pixel block, and obtain the pixel according to the direction value and the variance value. The classification result of the block, where the direction value contains 3 values, and the variance value contains 5 values. Taking a 4x4 pixel block as an example, referring to FIG. 2, the pixel classification method includes:

 Step 201: Calculate the horizontal activity vertical activity and the vertical activity horizontal activity of each 4x4 pixel block,

Vertical activity = sunii _d | (R(i,j )« 1 )-R(i- 1 ,j )-R(i+ 1 ,j ) | i,j =0...3

 Horizontal activity = sumij|(R(i,j)«l)-R(i,j-l)-R(i,j+l)| i,j=0...3

 Where i represents the relative vertical coordinate of the pixel, j represents the relative horizontal coordinate of the pixel; sum represents the summation operation; R represents the pixel

 Step 202: Obtain a direction value of the pixel block according to the horizontal activity and vertical activity of the pixel block; ^Vertical activity > threshold*Horizontal activity, ^ - define the direction value D = 1; ^口果Horizontal activity > threshold * Vertical activity, ^^ defines the direction value D = 2;

 In the remaining cases, the direction value D = 0.

 Step 203: Quantify the sum of the horizontal activity and the vertical activity of the 4x4 pixel block into 5 values as the variance value of the pixel block:

 A = Q (Horizontal activity + Vertical activity), Q ( - ) is a quantization function

 Step 204: Obtain a classification result of the pixel block according to the direction value and the variance value of the pixel block. The category C of the current 4x4 pixel block can be written as: C = A + 5 * D; where A is the gradient of the pixel and D is the direction value of the pixel.

In the process of implementing the present invention, the inventors have found that the prior art has at least the following disadvantages: This classification method utilizes the gradient of the local part of the pixel to represent the directional characteristic, and the local variance represents the texture intensity feature, although the different categories are distinguished to some extent. The pixels, but taking into account the main characteristics of the filter contain two parameters of bandwidth and direction, where the bandwidth reflects the strength of the airspace edge. However, the directional characteristics tend to have a more pronounced effect on the filter. The horizontal and vertical directions are used to distinguish the internal texture direction of the image, and the image characteristics are obviously not satisfied. Summary of the invention

 An object of the embodiments of the present invention is to provide a pixel classification and apparatus for video images, so as to improve the rationality of the pixel class arrangement method and improve the coding efficiency. According to an embodiment of the present invention, a pixel classification method for a video image is provided, including: obtaining a direction value corresponding to any pixel or a pixel block in a video image;

 Obtaining a complexity value corresponding to the pixel or the pixel block; wherein, the number of classifications of the direction value is greater than or equal to the number of classifications of the complexity value;

 The classification result of the pixel is obtained according to the direction value and the complexity value of the pixel. According to another embodiment of the present invention, a pixel classification apparatus for a video image is provided, including: a direction obtaining unit, configured to obtain a direction value corresponding to any pixel or a pixel block in a video image; a complexity obtaining unit, configured to Obtaining a complexity value corresponding to the pixel or the pixel block; wherein, the number of classifications of the direction value is greater than or equal to the number of classifications of the complexity value;

 And a classification unit, configured to obtain a classification result of the pixel according to the direction value and the complexity value of the pixel. The invention improves the proportion of the directional characteristic classification in the process of classifying the adaptively filtered pixels or pixel blocks, that is, the number of classifications of the direction values is greater than or equal to the number of classifications of the complexity values, thereby improving the rationality of the filter design. Finally, the coding efficiency is improved. DRAWINGS

 In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below. Obviously, the drawings in the following description are only It is a certain embodiment of the present invention, and other drawings can be obtained from those skilled in the art without any inventive labor.

 1 is a schematic structural view of a filter in the prior art;

 2 is a schematic flow chart of a pixel classification method in the prior art;

 FIG. 3 is a schematic flowchart diagram of an embodiment of a pixel classification method according to the present invention; FIG.

FIG. 4 is a schematic flowchart diagram of another embodiment of a pixel classification method according to the present invention; FIG. FIG. 5 is a schematic flowchart diagram of another embodiment of a pixel classification method according to the present invention; FIG. 6 is a schematic structural diagram of an embodiment of a pixel classification apparatus according to the present invention;

 FIG. 7 is a schematic structural diagram of another embodiment of a pixel sorting apparatus according to the present invention; FIG.

 FIG. 8 is a schematic structural diagram of another embodiment of a pixel sorting apparatus according to the present invention. Detailed ways

 BRIEF DESCRIPTION OF THE DRAWINGS The technical solutions in the embodiments of the present invention will be described in detail below with reference to the accompanying drawings. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative work are within the scope of the present invention.

 In the field of digital signal processing, video codecs are widely used in various electronic devices, such as: mobile phones, wireless devices, personal data assistants (PDAs), handheld or portable computers, GPS receivers/navigators, cameras, audio/ Video player, video camera, video recorder, monitoring equipment, etc. Generally, such an electronic device includes a video encoder or a video decoder, and the video encoder or video decoder can be directly implemented by a digital circuit or a chip such as a DSP (digital signal processor), or can be executed by a software code to execute the software in the processor. The process is implemented. The embodiment of the invention proposes a new pixel classification method, which not only considers the important influence of the pixel directional characteristic on the filter coefficient training, but also considers the influence of the local variance of the region where the pixel is located on the filter strength. In order to facilitate the combination of coefficients, a reasonable arrangement of pixel categories is designed, and finally the coding efficiency is improved. The directionality uses the gradient to calculate the direction and intensity of the local texture of the image, and the number of classifications of the direction values is greater than or equal to the number of classifications of the complexity values. Referring to FIG. 3, an embodiment of a pixel classification method for a video image provided by the present invention is used to obtain a classification of any pixel or pixel block in a video image, including:

 S301: Obtain a direction value corresponding to any pixel or pixel block in the video image;

 The pixel classification method can implement classification of one pixel or classification in units of pixel blocks.

The direction value corresponding to the pixel or pixel block is calculated using four gradient values. a step of this step The methods include:

 Calculating four gradient values corresponding to the pixel or pixel block: a horizontal gradient value, a vertical gradient value, a 45° angular gradient value, and a 135° angular gradient value;

 Obtaining a direction value corresponding to the pixel or the pixel block according to the four gradient values.

 Taking the class of pixels at ( , _ / ) and the gradient operator selection (-1, 2, -1) as an example, the four gradient values are calculated by the following formula, grad_h{i,j) X(i + k, j + l- 1)|

 Grad_v(i,j) = ^ *X(i + k,j + l)-X(i + k + \,j + l)\- X(i + k_l,j + /) grad_d(i,j) X(i + kl,j + 1 + grad_u(i,j) X(i + k + 1 -\)\

 Where grad—h(i,j), grad—v(i,j), grad—d^) and grad—u^) respectively represent (horizontal horizontal gradient, vertical gradient, 45° angular gradient, and 135° angular gradient) 4 gradient values, where X represents a pixel, i represents the vertical coordinate of the current pixel, j represents the horizontal coordinate of the current pixel, K represents the offset value of the current pixel vertical coordinate, and 1 represents the offset value of the current pixel horizontal coordinate.

 After obtaining four gradient values, an integer value of 4 values between 0 and 3 or an integer of 5 directions between 0 and 4 can be obtained from the four gradient values.

 S302: Obtain a complexity value corresponding to the pixel or the pixel block, where the number of classifications of the direction value is greater than or equal to the number of classifications of the complexity value;

 One embodiment of the step includes processing a gradient sum of the horizontal gradient value and the vertical gradient value using a quantization function to obtain a complexity value corresponding to the pixel or pixel block. Complexity values include: 0, 1, 2, 3 or 0, 1, 2.

 S303: Obtain a classification result of the pixel or the pixel block according to the direction value and the complexity value of the pixel or the pixel block.

The classification result C = A + 4 * D or the classification result C = A + 3 * D; where A is the complexity value of the pixel and D is the direction value of the pixel. The above method may be performed in the encoder or in the decoder. When in the decoder, the method further comprises: filtering the pixel or the pixel block by using the adaptive loop filter corresponding to the classification result.

 In the above embodiment, not only the important influence of the pixel directivity feature on the filter coefficient training is considered, but also the influence of the local variance of the region where the pixel is located on the filter strength is considered, and the number of the direction value is greater than or equal to the number of the complexity value. In order to facilitate the combination of coefficients, a reasonable arrangement of pixel categories is designed, and finally the coding efficiency is improved. Referring to FIG. 4, another embodiment of a pixel classification method for a video image provided by the present invention includes:

 S401: Calculate four gradient values corresponding to the pixel or the pixel block for any pixel or pixel block in the encoded image: horizontal gradient value, vertical gradient value, 45° angular gradient value, and 135° angular gradient value;

 Calculate the current local gradient feature of the pixel to calculate the class of the pixel at ( , _ / ·), and select the gradient operator (-1, 2, -1) as an example. Calculate the four gradients with the following formula: grad_h{i,j ) X(i + k,j + l- 1)|

 Grad_v(i,j) = ^ *X(i + k,j + l)-X(i + k + \,j + l)\- X(i + k_l,j + /) grad_d(i,j) X(i + kl,j + 1 + grad_u(i,j) X(i + k + 1 -\)\

 Where grad— h^ ), grad_v(y′), grad—d^ ) and grad— u^ ) represent the horizontal gradient of the (^) position, the vertical gradient, the 45° angular gradient and the 135° angular gradient, respectively. . Where X represents a pixel, i represents the vertical coordinate of the current pixel, j represents the horizontal coordinate of the current pixel, K represents the offset value of the current pixel vertical coordinate, and 1 represents the offset value of the current pixel horizontal coordinate.

 S402: Obtain a direction value corresponding to the pixel or the pixel block according to the four gradient values. Specifically, select a direction value corresponding to a minimum value of the four gradient values as a direction value of the pixel or the pixel block.

If the minimum of the four gradient values is a horizontal gradient, the direction value D = 0; If the minimum of the four gradient values is a 45 ^Q angle gradient, the direction value D = 1;

 If the minimum of the four gradient values is a vertical gradient, the direction value D = 2;

 If the minimum of the four gradient values is a 135° angular gradient, the direction value D = 3.

 S403: processing, by using a quantization function, a gradient sum of the horizontal gradient value and the vertical gradient value to obtain a complexity value corresponding to the pixel or the pixel block;

 The complexity value can reuse the horizontal gradient value and the vertical gradient value, and the gradient sum is quantized to an integer between 0 and 3; thus the variance at (i, j) is expressed as A = Q(grad_h(i,j) + grad_v (i, j)), where Q(') represents the quantization function, and the gradient is quantized to between 0-3.

 S404: Obtain a classification result of the pixel or the pixel block according to the direction value and the complexity value of the pixel or the pixel block.

 The final classification result of this pixel: C = A + 4 * D; where A is the gradient of the pixel and D is the direction value of the pixel.

 The foregoing method may be implemented on the decoding end and may also be implemented on the encoding end. Further, if the method is implemented on the decoding end, the method further includes the following steps:

 S405: Filter the pixel or the pixel block by using an ALF filter corresponding to the classification result.

ALF needs to pass some self-information compression coding to the decoding end. After receiving the code stream, the decoding end first needs to decode the information to obtain the ALF syntax element. And filtering the pixel or the pixel block by using the ALF filter corresponding to the classification result.

 In the above embodiment, the direction value of the pixel or the pixel block is obtained according to the four gradient values, the direction value includes 4 values, the complexity value includes 4 values, and the arrangement method of 16 pixel categories is obtained, and the texture directivity classification is greater than or equal to the texture. Complexity classification, this classification method can fully exploit the characteristics of pixels or pixel blocks, which is beneficial to the final merge operation of the filter, and finally achieves the improvement of coding efficiency. Referring to FIG. 5, another embodiment of a pixel classification method for a video image provided by the present invention includes:

 S501: Calculate four gradient values corresponding to the pixel or the pixel block for any pixel or pixel block in the encoded image: horizontal gradient value, vertical gradient value, 45° angular gradient value, and 135° angular gradient value;

Calculate the current pixel local gradient feature to calculate the class of pixels at ( , _ / · ), gradient operator Taking (-1, 2, -1) as an example, calculate four gradients using the following formula: grad_h{i,j) X(i + k,j + l- 1)|

 Grad_v(i,j) = ^ *X(i + k,j + l)-X(i + k + \,j + l)\- X(i + k_l,j + /) grad_d(i,j) X(i + kl,j + 1 + grad_u(i,j) X(i + k + 1 -\)\

 Where grad— h^ ), grad_v(y′), grad—d^ ) and grad— u^ ) represent the horizontal gradient of the (^) position, the vertical gradient, the 45° angular gradient and the 135° angular gradient, respectively. .

 S502: Obtain a direction value corresponding to the pixel or the pixel block according to the four gradient values. If the absolute value of the difference between the maximum value and the minimum value of the four gradient values is less than a threshold value, the direction value is the first direction. Value; otherwise, the direction value corresponding to the minimum of the four gradient values is selected as the direction value of the pixel or the pixel block;

 If the absolute value of the difference between the minimum gradient value and the maximum gradient value is less than a threshold value, the direction value D=0;

 Otherwise, the direction value corresponding to the minimum value among the four gradient values is selected as the direction value of the pixel: if the minimum value among the four gradient values is a horizontal gradient, the direction value D = 0;

 If the minimum of the four gradient values is a 45° angular gradient, the direction value D = 1;

 If the minimum of the four gradient values is a vertical gradient, the direction value D = 2;

 If the minimum of the four gradient values is a 135° angular gradient, the direction value D = 3.

 S503: processing, by using a quantization function, a gradient sum of the horizontal gradient value and the vertical gradient value to obtain a complexity value corresponding to the pixel or the pixel block;

 The complexity value can reuse the horizontal gradient value and the vertical gradient value, and the gradient sum is quantized to an integer between 0 and 2; thus the variance at (i, j) is expressed as A = Q(grad_h(i,j) + grad_v (i, j)), where Q(') represents the quantization function, and the gradient is quantized to between 0-2.

S504: Obtain a classification result of the pixel according to the direction value and the complexity value of the pixel; and finally classify the result of the pixel: C=A+3*D; where A is a gradient of the pixel and D is a direction of the pixel. value. The foregoing method may be implemented on the decoding end and may also be implemented on the encoding end. Further, if the method is implemented on the decoding end, the method further includes the following steps:

 S505: Filter the pixel or the pixel block by using an ALF filter corresponding to the classification result. ALF needs to pass some self-information compression coding to the decoding end. After receiving the code stream, the decoding end first needs to decode the information to obtain the ALF syntax element. And filtering the pixel or the pixel block by using the ALF filter corresponding to the classification result.

 In the above embodiment, the direction value of the pixel or the pixel block is obtained according to the four gradient values, the direction value includes five values, the complexity value includes three values, and the arrangement method of the 15 pixel categories is obtained, and the texture directivity classification is greater than or equal to the texture. Complexity classification, this classification method can fully exploit the characteristics of pixels or pixel blocks, which is beneficial to the final merge operation of the filter, and finally achieves the improvement of coding efficiency. A person skilled in the art can understand that all or part of the process of implementing the above embodiment method can be completed by a computer program to instruct related hardware, and the program can be stored in a computer readable storage medium, the program When executed, the flow of an embodiment of the methods as described above may be included. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM). In association with the above method embodiments, the present invention also provides a pixel classification apparatus for a video image, which may be located in an encoder or a decoder. The pixel classifying device of the video image may be implemented by a hardware circuit or by software in cooperation with hardware. For example, referring to Figure 6, a pixel classification device that calls a video image by a processor implements classification of pixels. The pixel classifying device of the video image can perform various methods and processes in the above method embodiments. Referring to FIG. 7, an embodiment of a pixel classification apparatus for a video image according to the present invention includes: a direction obtaining unit 701, configured to obtain a direction value corresponding to any pixel or a pixel block in a video image;

 The complexity obtaining unit 702 is configured to obtain a complexity value corresponding to the pixel or the pixel block, where the number of classifications of the direction value is greater than or equal to the number of classifications of the complexity value;

a classifying unit 703, configured to obtain a classification of the pixel according to the direction value and the complexity value of the pixel The result. Referring to FIG. 8, an embodiment of a pixel classifying apparatus for a video image according to the present invention includes: a gradient value calculating unit 801, configured to calculate four gradient values corresponding to the pixel or the pixel block: a horizontal gradient value, a vertical gradient value, and 45 ^Q angle gradient value and 135° angle gradient value;

 The obtaining unit 802 is configured to obtain, according to the four gradient values, a direction value corresponding to the pixel or the pixel block;

 a complexity obtaining unit 803, configured to process a gradient sum of the horizontal gradient value and the vertical gradient value by using a quantization function, to obtain a complexity value corresponding to the pixel or the pixel block; wherein, the number of classifications of the direction value is greater than or equal to the complexity The number of categories of values;

 The classifying unit 804 is configured to obtain a classification result of the pixel according to the direction value and the complexity value of the pixel.

 In one embodiment, the complexity obtaining unit 803 selects a direction value corresponding to a minimum value among the four gradient values as a direction value of the pixel or the pixel block.

 Correspondingly, the horizontal gradient value corresponds to the direction value D = 0; 45° angular gradient value corresponds to the direction value 0 =

1; the direction value corresponding to the vertical gradient value D = 2; the direction value corresponding to the 135° angular gradient D = 3; the complexity values include: 0, 1 , 2, 3.

 The classification result C = A + 4 * D; where A is the complexity value of the pixel and D is the direction value of the pixel. In another embodiment, if the absolute value of the difference between the maximum value and the minimum value of the four gradient values is less than a threshold value, the complexity obtaining unit 803 determines that the direction value corresponding to the pixel or the pixel block is the first direction value. Otherwise, the complexity obtaining unit selects a direction value corresponding to a minimum value among the four gradient values as a direction value of the pixel, wherein the first direction value is a value different from a direction value corresponding to the four gradient values.

 Correspondingly, the first direction value D = 0; the horizontal gradient value corresponds to the direction value D = 1; the 45° angular gradient value corresponds to the direction value D = 2; the vertical gradient value corresponds to the direction value D = 3; 135 ° angular gradient The corresponding direction value D = 4; complexity values include: 0, 1 , 2.

Classification result = eight + 3 * 0; where A is the complexity value of the pixel and D is the direction value of the pixel. The pixel classification device of the above video image may be implemented in an encoder or may be implemented in a decoder. When the pixel classification device of the video image is applied to the encoding end, the method further includes: And a filtering module, configured to filter the pixel or the pixel block by using an adaptive loop filter corresponding to the classification result.

 The above embodiment not only considers the important influence of the pixel directivity feature on the filter coefficient training, but also considers the influence of the local variance of the pixel region on the filter strength. The number of the direction value is greater than or equal to the number of the complexity value. A reasonable arrangement of pixel categories is used to achieve an improvement in coding efficiency.

 The above is only a few embodiments of the present invention, and those skilled in the art can make various changes or modifications to the invention without departing from the spirit and scope of the invention.

+

Claims

Rights request

 A pixel classification method for a video image, characterized in that:

 Obtaining a direction value corresponding to any pixel or pixel block in the video image;

 Obtaining a complexity value corresponding to the pixel or the pixel block; wherein, the number of classifications of the direction value is greater than or equal to the number of classifications of the complexity value;

 The classification result of the pixel is obtained according to the direction value and the complexity value of the pixel.

 2. The method according to claim 1, wherein obtaining a direction value corresponding to any pixel or pixel block in the video image comprises:

 Calculating four gradient values corresponding to the pixel or pixel block: a horizontal gradient value, a vertical gradient value, a 45° angular gradient value, and a 135° angular gradient value;

 Obtaining a direction value corresponding to the pixel or the pixel block according to the four gradient values;

 The obtaining the complexity value corresponding to the pixel or the pixel block includes:

 A gradient sum of the horizontal gradient value and the vertical gradient value is processed by a quantization function to obtain a complexity value corresponding to the pixel or pixel block.

 The method according to claim 2, wherein, according to the four gradient values, obtaining a direction value corresponding to the pixel or the pixel block comprises:

 A direction value corresponding to a minimum value among the four gradient values is selected as a direction value of the pixel or the pixel block.

The method according to claim 3, wherein the horizontal gradient value corresponds to a direction value D = 0; the 45° angular gradient value corresponds to a direction value D = 1; The direction value D = 2; the direction value corresponding to the 135 ^Q angle gradient is D = 3; the complexity value includes: 0, 1 , 2, 3.

 The method according to claim 4, wherein the classification result of the pixel is obtained according to the direction value and the gradient sum of the pixel: the classification result=eight+4*0; wherein A is a complexity value of the pixel , D is the direction value of the pixel.

 The method according to claim 2, wherein, according to the four gradient values, obtaining a direction value corresponding to the pixel or the pixel block comprises:

If the absolute value of the difference between the maximum value and the minimum value of the four gradient values is less than a threshold value, determining that the direction value corresponding to the pixel or the pixel block is the first direction value; otherwise, selecting the four gradient values The direction value corresponding to the medium minimum value is taken as the direction value of the pixel, wherein the first direction value is the same as the four gradient values The corresponding direction value is different.

 The method according to claim 6, wherein the first direction value D = 0; the horizontal gradient value corresponding to the direction value D = 1; the 45° angular gradient value corresponding to the direction value D = 2; the vertical gradient value corresponds to a direction value D = 3; the 135 ° angular gradient corresponds to a direction value D = 4; the complexity value includes: 0, 1 , 2.

 The method according to claim 7, wherein the classification result of the pixel is obtained according to the direction value and the complexity value of the pixel: the classification result=eight+3*0; wherein A is the complexity of the pixel Value, D is the direction value of the pixel.

 9. The method according to claim 1, further comprising:

 The pixel or pixel block is filtered using an adaptive loop filter corresponding to the classification result.

10. A pixel classification device for video images, comprising:

 a direction obtaining unit, configured to obtain a direction value corresponding to any pixel or a pixel block in the video image; a complexity obtaining unit, configured to obtain a complexity value corresponding to the pixel or the pixel block; wherein, the number of the direction value is greater than Or the number of classifications equal to the complexity value;

 And a classification unit, configured to obtain a classification result of the pixel according to the direction value and the complexity value of the pixel.

The apparatus according to claim 10, wherein the direction obtaining unit comprises: a gradient value calculating unit, configured to calculate four gradient values corresponding to the pixel or the pixel block: a horizontal gradient value, a vertical gradient value , 45 ^Q angle gradient values and 135 ^Q angle gradient values;

 And an obtaining unit, configured to obtain, according to the four gradient values, a direction value corresponding to the pixel or the pixel block;

 The complexity obtaining unit is configured to process the gradient sum of the horizontal gradient value and the vertical gradient value by using a quantization function to obtain a complexity value corresponding to the pixel or the pixel block.

 The apparatus according to claim 11, wherein the complexity obtaining unit selects a direction value corresponding to a minimum value among the four gradient values as a direction value of the pixel or the pixel block.

 The device according to claim 12, wherein the horizontal gradient value corresponds to a direction value D = 0; the 45° angular gradient value corresponds to a direction value D = 1; The direction value D = 2; the 135° angular gradient corresponds to a direction value D = 3; the complexity value includes: 0, 1 , 2, 3.

14. The apparatus according to claim 13, wherein the classification result C = A + 4 * D; wherein A is a complexity value of the pixel, and D is a direction value of the pixel.

15. The apparatus according to claim 11, wherein the complexity obtaining unit determines the pixel or pixel if an absolute value of a difference between a maximum value and a minimum value among the four gradient values is less than a threshold value The direction value corresponding to the block is a first direction value; otherwise, the complexity obtaining unit selects a direction value corresponding to a minimum value among the four gradient values as a direction value of the pixel, where the first direction value is four gradients The value corresponding to the direction value is different.

 The device according to claim 15, wherein the first direction value D = 0; the horizontal gradient value corresponds to a direction value D = 1; and the 45° angular gradient value corresponds to a direction value D = 2; the vertical gradient value corresponds to a direction value D = 3; the 135 ° angular gradient corresponds to a direction value D = 4; the complexity value includes: 0, 1 , 2.

 17. Apparatus according to claim 16 wherein said classification result C = A + 3

* D; where A is the complexity value of the pixel and D is the direction value of the pixel.

 The device according to claim 10, further comprising:

 And a filtering module, configured to filter the pixel or the pixel block by using an adaptive loop filter corresponding to the classification result.