CN101552917B

CN101552917B - Bit rate control method for video compression

Info

Publication number: CN101552917B
Application number: CN 200810066409
Authority: CN
Inventors: 徐苏珊
Original assignee: SHENZHEN RONGCHUANG TIANXIA TECHNOLOGY DEVELOPMENT Co Ltd
Current assignee: Shenzhen Temobi Science and Technology Co Ltd
Priority date: 2008-03-31
Filing date: 2008-03-31
Publication date: 2011-01-05
Anticipated expiration: 2028-03-31
Also published as: CN101552917A; WO2009121234A1

Abstract

The invention relates to the field of video image processing and provides a bit rate control method for video compression. The method comprises the following steps: A. image features are analyzed, andimage feature complexity factors are calculated; B. the number of target bits which are needed to be distributed to an image is calculated; and C. quantization parameters of codes are obtained by a r ate distortion model and combining the number of target bits needed to be distributed to the image. In the invention, the bit rate distribution and the image feature complexity are closely combined, and the quantization parameters Q p of codes are obtained by a secondary non-linear rate distortion model, thereby increasing the bit rate control accuracy and improving the image quality.

Description

Video compression code rate control method

Technical Field

The invention relates to the field of video image processing, in particular to a video compression code rate control method.

Background

In the video compression process, the data amount of the compressed frame varies, so that the data amount per unit time (code rate) fluctuates. In practical applications, especially in video streaming, a constant bitrate is required. The principle of code rate control is to utilize a mathematical model to recalculate the quantization factor of the next frame by the currently known user-specified code rate and the compressed bit number, thereby changing the bit number after encoding and achieving the purpose of controlling the code rate.

A conventional rate control algorithm (e.g., CBR algorithm) includes the following steps: (1) allocating a target number of bits almost equally to each frame image according to a target bit rate; (2) and calculating the encoded quantization parameter Qp for each frame according to the allocated target bit number, thereby ensuring that the encoder outputs a constant bit rate. Because the algorithm of the prior art allocates almost the same bit number to each frame of image in the video sequence, and actually, the complexity of each frame of image is constantly changed, the quality of the compressed image has great fluctuation, so that the quality of the image is not high, and especially in the video compression under a low bandwidth/wireless channel, because the channel has the characteristics of instability and error easiness, higher requirements are provided for the code rate control precision of an encoder.

Therefore, a new video compression rate control algorithm is needed to improve the rate control accuracy and thus improve the image quality.

Disclosure of Invention

The invention aims to provide a video compression code rate control algorithm, which aims to solve the problems of low code rate control precision and large image quality fluctuation in the video compression process in the prior art.

In order to achieve the object of the invention, the video compression rate control algorithm comprises the following steps:

A. analyzing the image characteristics and calculating the complexity factor of the image characteristics;

B. calculating the target bit number to be distributed by the image by combining the image characteristic complex factor;

C. and (4) calculating a coding quantization parameter QP by using a rate distortion model according to the target bit number required to be allocated to the image.

Preferably, the step a further comprises:

A1. calculating the motion complexity factor of the current image according to the actual bit number and the output average bit number generated by image coding;

A2. calculating the texture complexity factor of the current image according to the average texture complexity of the image;

A3. and combining the current motion complexity factor and the texture complexity factor of the image to calculate the characteristic complexity factor of the current image.

Preferably, the motion complexity factor of the image in step a1 is calculated by the formula:

C_{m} = \frac{B_{j}}{{\tilde{B}}_{j}}

wherein, C_mIs a motion complexity factor, B, of the current picture_jIs the actual number of bits generated by the coding of the image of the jth frame,

is the average number of bits of the encoded output calculated up to the j-th frame image.

Preferably, the

The calculation formula of (2) is as follows:

wherein,

is the average number of coded output bits calculated up to the j-th frame image,

is the average number of bits of the encoded output up to the j-1 th frame image, B_jIs the actual bit number of the coded output of the j frame image, and alpha is the weighting coefficient.

Preferably, the texture complexity factor of the image in step a2 is calculated by the formula:

C_{t} = \frac{\log_{2} ({MAD}_{j}) / M}{{\tilde{Ct}}_{j}}

wherein, C_tIs a texture complexity factor, MAD, of an image_jIs the average of the absolute differences of the previous picture, M is the number of macroblocks in the picture,

is the average texture complexity factor up to the jth frame image.

Preferably, theThe calculation formula of (2) is as follows:

wherein,

is the mean texture complexity factor, MAD, until the jth frame image_jIs the average of the absolute differences of the previous picture,

is the average texture complexity factor up to the j-1 frame image, and β is the addition coefficient.

Preferably, the calculation formula of the feature complexity factor of the image in the step a3 is:

C_j＝C_m+η×C_t

wherein, C_jIs a characteristic complexity factor of the image, C_mIs a motion complexity factor, C, of the image_tIs the texture complexity factor of the image and η is the adjustment factor.

Preferably, the target number of bits to be allocated for the image in step B is calculated by the following formula:

wherein, B_(n，j)Is the target number of bits to be allocated for the j frame image of the n second, C_(n，j)Is a feature complexity factor, R, of the corresponding image, i.e. the image of the j frame in the n second_bitIs the target bit rate of image coding, f_sIs the target frame rate, br, of the image coding_(i，m)Is the actual number of bits of the ith second mth picture output, and α isAnd adjusting the coefficient.

Preferably, said step C further comprises using a quadratic non-linear rate distortion model to find the encoded quantization parameter Qp in combination with the target number of bits to be allocated to the picture.

Preferably, the second order nonlinear rate distortion model is:

where B is the target number of bits allocated for the current frame, SAD is the sum of absolute differences of the current frame, Qp is the quantization parameter for the macroblock, and c1 and c2 are the adjustment parameters.

The invention closely combines the code rate distribution with the characteristic complexity of the image, so that the encoder can accurately distribute the code rate and improve the control precision of the code rate; meanwhile, the invention uses a quadratic nonlinear rate distortion model to obtain the quantization parameter Qp of the code, and compared with a simple linear rate distortion model, the accuracy of controlling the code rate is better. Therefore, the video compression code rate control algorithm of the invention can improve the code rate control precision, thereby improving the image quality.

Drawings

FIG. 1 is a flow chart of a video compression rate control method according to the present invention;

fig. 2 is a flowchart of a video compression rate control method according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

The invention combines the code rate distribution and the image characteristic complexity to distribute the code rate, and then obtains the coding quantization parameter through the secondary nonlinear rate distortion model, thereby improving the code rate control precision and improving the image quality.

Fig. 1 shows a flow chart of the video compression rate control method of the present invention, and the process is as follows:

in step S101, the image features are analyzed and the image feature complexity factor is calculated.

In step S102, the target number of bits to be allocated to the image is calculated by combining the complexity factor of the image feature.

In step S103, the rate-distortion model is used to obtain the quantization parameter QP for coding, in combination with the target number of bits to be allocated to the picture.

Fig. 2 shows a flowchart of a video compression rate control method according to an embodiment of the present invention, which is based on the method flow shown in fig. 1, and includes the following specific processes:

in step S201, a motion complexity factor of a current picture is calculated based on an actual number of bits generated by picture coding and an output average number of bits.

The motion complexity of an image refers to the difference caused by the motion of an object in two consecutive frames of images in an image sequence, and the correlation between two consecutive frames of images in the image sequence is very large. One embodiment of the invention analyzes the motion complexity of the image through the previous frame image, and expresses the motion complexity by using the motion complexity factor of the image. In one embodiment, the motion complexity factor for the current image is calculated by the formula:

C_{m} = \frac{B_{j}}{{\tilde{B}}_{j}}

wherein, C_mIs a motion complexity factor, B, of the current picture_jIs the actual number of bits generated by the coding of the image of the jth frame,is the average number of bits of the encoded output calculated up to the j-th frame image.

In one embodiment, the average number of bits of the encoded output calculated up to the jth frame image

The calculation formula of (2) is as follows:

wherein,

is the average number of bits of the encoded output up to the j-1 th frame image, B_jIs the actual bit number of the coded output of the j frame image, and alpha is the weighting coefficient. The value of α may be adjusted according to specific situations, and in one embodiment, the value α is 0.725.

In step S202, a texture complexity factor of the current image is calculated based on the average texture complexity of the image. Since the scene correlation of two consecutive frames of images is relatively large in the image sequence, one embodiment of the present invention uses the previous frame of image to analyze the texture complexity of the image, and uses the texture complexity factor of the image to express the texture complexity. In one embodiment, the texture complexity factor for the current image is calculated as:

C_{t} = \frac{\log_{2} ({MAD}_{j}) / M}{{\tilde{Ct}}_{j}}

is the average texture complexity factor up to the jth frame image.

In one embodiment, the average texture complexity factor up to the jth frame image

The calculation formula of (2) is as follows:

wherein,is the mean texture complexity factor, MAD, until the jth frame image_jIs the average of the absolute differences of the previous picture,

is the average texture complexity factor up to the j-1 frame image, and β is the addition coefficient. The value of β may be adjusted according to specific situations, and in one embodiment, β may be 0.825.

In step S203, the feature complexity factor of the current image is calculated by combining the current motion complexity factor and texture complexity factor of the image. One embodiment of the present invention uses the feature complexity factor of the image to describe the image feature, and combines the motion complexity factor and the texture complexity factor of the image to calculate the feature complexity factor of the image, and the calculation formula is as follows:

C_j＝C_m+η×C_t

wherein, C_jIs a characteristic complexity factor of the image, C_mIs a motion complexity factor, C, of the image_tIs the texture complexity factor of the image and η is the adjustment factor. The value of η may be adjusted according to specific situations, and in one embodiment, η may be equal to 0.5.

In step S204, the image feature is introduced into the image bit rate allocation process, and the target number of bits to be allocated to the image is calculated by combining the complexity factor of the image feature. In one embodiment, the target number of bits to be allocated for a picture is calculated as:

wherein, B_(n，j)Is the target number of bits to be allocated for the j frame image of the n second, C_(n，j)Is a feature complexity factor, R, of the corresponding image, i.e. the image of the j frame in the n second_bitIs the target bit rate of image coding, f_sIs the target frame rate, br, of the image coding_(i，m)Is the actual number of bits output by the mth picture in the ith second, and α is the adjustment coefficient. The value of α may be adjusted according to specific situations, and in one embodiment, the value α is 0.25. It should be noted that the above calculation formula is only one example of the present invention, and other transformation formulas of the formula are also included in the protection scope of the present invention.

In step S205, a Rate-Distortion (RD) model is used to determine the encoded quantization parameter Qp according to the target number of bits to be allocated to the picture. In an exemplary embodiment of the present invention, the quantization parameter Qp of the code is obtained using a quadratic non-linear RD model in combination with the target number of bits to be allocated to the picture, given the target number of bits already known.

In one embodiment, the quadratic non-linear RD model used is:

where B is the target number of bits allocated for the current frame, SAD (Sum of Absolute Difference) is the Sum of Absolute differences of the current frame, Qp is the quantization parameter of the macroblock, and c1 and c2 are the adjustment parameters. Where c1 and c2 are estimated from the previous frame image using the RD model and need to be updated continuously.

The quantization parameter Qp, which is an important parameter for the encoder to control the degree of image compression, is used to control the quantizer in encoding, and the smaller the Qp, the finer the quantization, the higher the image quality, and the longer the generated code stream. The secondary nonlinear RD model is used for dynamically changing the quantization parameter Qp, so that the complexity of an input image and the output code rate can be balanced, and the code rate of the coded output is constant.

It should be noted that the present invention is a generic algorithm that can be applied to different encoders, such as any one of the h.120, h.261, h.263, h.264, MPEG-1, MPEG-4 or any other hybrid framework encoders.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.

Claims

1. A method for controlling video compression rate, the method comprising the steps of:

C. calculating a coding quantization parameter QP by using a rate distortion model in combination with the target bit number to be distributed to the image;

the step A further comprises the following steps:

2. The method for controlling video compression rate according to claim 1, wherein the motion complexity factor of the picture in step a1 is calculated by the formula:

C_{m} = \frac{B_{j}}{{\tilde{B}}_{j}}

3. The method of claim 2, wherein the video compression rate control method is further characterized in that

The calculation formula of (2) is as follows:

wherein,

4. The method for rate control in video compression as claimed in claim 1, wherein the texture complexity factor of the picture in step a2 is calculated by the formula:

C_{t} = \frac{\log_{2} ({MAD}_{j}) / M}{\tilde{C} t_{j}}

is the average texture complexity factor up to the jth frame image.

5. The method of claim 4, wherein the video compression rate control method is applied to the video compressionThe calculation formula of (2) is as follows:

wherein,

is the average texture complexity factor until the j-1 frame image, and β is the weighting factor.

6. The method for controlling video compression rate according to claim 1, wherein the calculation formula of the characteristic complexity factor of the image in step a3 is:

C_j＝C_m+η×C_t

7. The method of claim 1, wherein the target number of bits to be allocated for the picture in step B is calculated by the following formula:

wherein, B_(n，j)Is the target number of bits to be allocated for the j frame image of the n second, C_(n，j)Is a feature complexity factor, R, of the corresponding image, i.e. the image of the j frame in the n second_bitIs the target bit rate of image coding, f_sIs the target frame rate, br, of the image coding_(i，m)Is the actual number of bits output by the mth picture in the ith second, and α is the adjustment coefficient.

8. The method of claim 1, wherein step C further comprises using a quadratic non-linear rate distortion model to obtain the encoded quantization parameter Qp in combination with the target number of bits to be allocated for the picture.

9. The method of claim 8, wherein the quadratic non-linear rate distortion model is: