CN100546388C - A kind of chroma equalization method that is used for video coding - Google Patents

Abstract

The present invention relates to a kind of chroma equalization method and chromaticity quantization parameter correction method that is used for video coding, this equalization methods comprises: obtain the luminance component residual error SAD between present image and the predicted picture _YWith chromatic component residual error SAD _CObtaining luminance component residual error SAD _YWith chromatic component residual error SAD _CAfter calculate residual error distribution SATD in the frequency domain _YAnd SATD _CBased on above-mentioned residual error distribution SAD _Y, SATD _CAnd luminance quantization parameter QP _YRevise chromaticity quantization parameter QP _CRespectively based on luminance quantization parameter QP _YWith chromaticity quantization parameter QP _CLuminance component and chromatic component to present image are encoded.Pass through said method, utilize the comparison of present image brightness and chromatic component residual error and distribution thereof, regulate the chromaticity quantization parameter, the compression ratio of control chromatic component, thereby the Y-PSNR of balance chromatic component and luminance component, so balance the two reconstruction quality and save code check.

Description

A kind of chroma equalization method that is used for video coding

[technical field]

The present invention relates to a kind of method for video coding, be specifically related to a kind of chroma equalization method and chromaticity quantization parameter correction method that is used for video coding.

[background technology]

At the real-time video communication situation, the sample format of vision signal is Y:U:V=4:2:0 normally, and digital video coding standard Baseline Profile (basic class) H.264 also only supports the Y:U:V=4:2:0 form.As shown in Figure 1, under this sample format, a sampled point of the average per corresponding chromatic components of four sampled points of luminance component (Y) (U, V).Why adopting this sample format is the characteristic of having utilized human vision, promptly to the sensitivity of luminance component greater than sensitivity to chromatic component.

Luminance component and chromatic component are in different their differences on the image encoding characteristic that determined on the spatial sampling rate.In the interframe encode process, digital video coding standard H.264 grammatical norm is only carried out motion search to luminance component, and the motion vector of chromatic component directly is taken from the luminance component of corresponding region.In the sub-pix territory, because the difference of spatial sampling rate, to the motion vector of same precision, chromatic component must be done interpolation one time by the specific luminance component more, and this has caused chromatic component distribution spatially level and smooth relatively.In addition, H.264 introduced the motion search of 1/4 pixel precision, luminance component is when making 1/4 picture element interpolation, chromatic component is to 1/8 picture element interpolation, this has more aggravated chromatic component flatness spatially, thereby cause the residual values of chromatic component less than normal, make usually specific luminance component height of the PSNR (Y-PSNR) of chromatic component after quantification.Test encoding softwares to two standard test sequences Mother﹠amp from some; PSNR value behind Daughter and the Highway Drive coding finds out, inter coded images is not had the compression ratio of equilibrium brightness, chromatic component, causes the PSNR value of entire image unbalanced.Can see, the high 0.4～1dB of the common specific luminance component of the PSNR of chromatic component, this has caused the imbalance of compression.Yet owing to human subjective effect, the PSNR specific luminance component height of chromatic component there is no need too much under normal conditions, and can not produce too big improvement to the quality of image, can waste suitable code check on the contrary.

[summary of the invention]

In order to solve in the prior art in the image encoding process because the spatial sampling rate difference of brightness and chromatic component, the compression imbalance that causes brightness and chromatic component, waste the technical problem of suitable code check, the invention provides a kind of chroma equalization method that is used for video coding and chromaticity quantization parameter correction method of can self adaptation regulating the chromatic component compression ratio.

The present invention solves in the prior art in the image encoding process because the spatial sampling rate difference of brightness and chromatic component, the compression imbalance that causes brightness and chromatic component, the technical scheme that technical problem adopted of having wasted suitable code check is: a kind of chroma equalization method that is used for video coding is provided, and this equalization methods comprises: a. obtains the luminance component residual error SAD between present image and the predicted picture _YWith chromatic component residual error SAD _CB. based on obtaining luminance component residual error SAD _YWith chromatic component residual error SAD _CCalculate the residual error distribution SATD in the frequency domain again _YAnd SATD _CC. based on residual error distribution SATD _Y, SATD _CAnd the luminance quantization parameter QP that obtains in advance _YDraw chromaticity quantization parameter QP by the formula correction _CD. respectively based on luminance quantization parameter QP _YWith chromaticity quantization parameter QP _CLuminance component and chromatic component to present image are encoded.

According to one preferred embodiment of the present invention, in step c, chromaticity quantization parameter Qp _CDraw by following formula:

${\mathrm{Qp}}_C=Q{p}_Y+\mathrm{\ΔQp}=Q{p}_Y+\mathrm{int}(\frac{1}{Q{p}_Y/\mathrm{\ω}}+(\frac{{\mathrm{SATD}}_Y}{N_Y}-\frac{{\mathrm{SATD}}_C}{N_C})\·\mathrm{\σ}),$

Wherein, int () represents bracket function, N _YExpression luminance pixel quantity, N _CExpression chroma pixel quantity, σ and ω are modifying factor.

According to one preferred embodiment of the present invention, the σ value is 0.4～1, and ω is 4～10.

According to one preferred embodiment of the present invention, in step b, residual error distribution SATD _Y, SATD _CThe Hamilton conversion approximate by integer transform obtains.

According to one preferred embodiment of the present invention, the sample format of present image and predicted picture is YUV4:2:0.

According to one preferred embodiment of the present invention, chromatic component comprises U chromatic component or V chromatic component.

The present invention solves in the prior art in the image encoding process because the spatial sampling rate difference of brightness and chromatic component, the compression imbalance that causes brightness and chromatic component, another technical scheme that technical problem adopted of having wasted sizable code check is: a kind of chromaticity quantization parameter correction method that is used for video coding is provided, comprises: a. obtains luminance component and the residual error distribution SATD of chromatic component in frequency domain _YAnd SATD _CAnd luminance quantization parameter QP _YB. obtain chromaticity quantization parameter Qp by following formula _C:

${\mathrm{Qp}}_C={\mathrm{Qp}}_Y+\mathrm{\ΔQp}=Q{p}_Y+\mathrm{int}(\frac{1}{{\mathrm{Qp}}_Y/\mathrm{\ω}}+(\frac{{\mathrm{SATD}}_Y}{N_Y}-\frac{{\mathrm{SATD}}_C}{N_C})\·\mathrm{\σ}),$

Wherein: int () represents bracket function, N _YExpression luminance pixel quantity, N _CExpression chroma pixel quantity, σ and ω are modifying factor.

According to one preferred embodiment of the present invention, σ is 0.4～1, and ω is 4～10.

According to one preferred embodiment of the present invention, in step b, residual error distribution SATD _Y, SATD _CThe Hamilton conversion approximate by integer transform obtains.

By said method, utilize the comparison of present image brightness and chromatic component residual error and distribution thereof, regulate chromaticity quantization parameter Qp _CThereby the compression ratio of control chromatic component, the Y-PSNR of balance chromatic component and luminance component, and then balance the two reconstruction quality and save code check, can see that the PSNR value of chromatic component is near luminance component, each component equiblibrium mass distribution of entire image, and saved suitable code check, by the observation to image, subjective quality difference is controlled at the degree that human eye is difficult to perceive.

[description of drawings]

Fig. 1 is the pixel distribution schematic diagram of existing video signal sample format YUV4:2:0;

Fig. 2 is the schematic flow sheet of chroma equalization method of the present invention;

Fig. 3 adopts existing JM encoding software and adds the comparison diagram as a result that balance method JM encoding software of the present invention is encoded to standard test sequences Silent;

Fig. 4 adopts existing JM encoding software and adds the comparison diagram as a result that balance method JM encoding software of the present invention is encoded to standard test sequences Foreman.

[embodiment]

Referring to Fig. 2, Fig. 2 is the schematic flow sheet of chroma equalization method of the present invention.In the present invention, by analyzing present image brightness and chromatic component residual error and distribution thereof, predict the luminance component PSNR of reconstructed image and the difference of chromatic component PSNR, and then by revising the chromaticity quantization parameter, thereby the compression ratio of control chromatic component, this equalization methods may further comprise the steps:

1. obtain the luminance component residual error SAD between present image and the predicted picture _YWith chromatic component residual error SAD _C

The residual values of brightness, chromatic component can be calculated by following formula:

$\mathrm{SAD}=\underset{i,j}{\mathrm{\Σ}}\left|\mathrm{Diff}(i,j)\right|,$ Diff(i，j)＝Original(i，j)-Prediction(i，j) (1)

Wherein, SAD (Sum of Absolute Difference, absolute difference and), ((i j) represents the brightness value or the chromatic value of the predicted picture behind motion search respectively to Prediction to Original for i, j) brightness value or the chromatic value of expression original image.

2. calculate luminance component residual error SAD _YWith chromatic component residual error SAD _CResidual error distribution SATD in frequency domain _YAnd SATD _C

With SAD _YAnd SAD _CCarry out conversion, draw the distribution (because quantize in frequency domain carry out) of above-mentioned residual error in frequency domain

In the present embodiment,, select Hamilton (Hadamard) conversion: the DiffT=H * Diff * H approximate for use with integer transform in order to reduce amount of calculation, wherein:

$\mathrm{SATD}=\left(\underset{i,j}{\mathrm{\Σ}}\right|\mathrm{DiffT}(i,j)\left|\right)/2---\left(2\right)$

SATD (Sum of Absolute Transform Difference, after the conversion absolute difference and) is illustrated in that residual error distributes in the frequency domain.

3. based on residual error distribution SATD _Y, SATD _CAnd luminance quantization parameter QP _YRevise chromaticity quantization parameter QP _C

Utilize above-mentioned steps 1,2 to calculate the SATD of luminance component and chromatic component respectively _YAnd SATD _CWherein Y is a luminance component, and C can be chromatic component U or V chromatic component.Utilize luminance quantization parameter QP _YBy the revised chromaticity quantization parameter of following formula QP _C

${\mathrm{Qp}}_C=Q{p}_Y+\mathrm{\ΔQp}=Q{p}_Y+\mathrm{int}(\frac{1}{Q{p}_Y/\mathrm{\ω}}+(\frac{{\mathrm{SATD}}_Y}{N_Y}-\frac{{\mathrm{SATD}}_C}{N_C})\·\mathrm{\σ}),$

Wherein, int () represents bracket function, N _YExpression luminance pixel quantity, N _CExpression chroma pixel quantity, σ and ω are the modifying factor of formulating according to experimental observation, generally, σ value 0.6, ω value 6.

4. respectively based on luminance quantization parameter QP _YWith chromaticity quantization parameter QP _CLuminance component and chromatic component to present image are encoded.

Concrete coding step is the known processes of this area, generally comprises orthogonal transform, quantification and entropy coding, and does not repeat them here.

By above-mentioned balance method, utilize the comparison of present image brightness and chromatic component residual error and distribution thereof, regulate the chromaticity quantization parameter, can control the compression ratio of chromatic component, thereby the Y-PSNR of balance chromatic component and luminance component, so balance the two reconstruction quality and save code check.

Shown in following table 1 and table 2, the corresponding data of " original series PSNR " in table 1 and the table 2 is respectively that the official of JVT (joint video code set) issue tests encoding software JM to two standard test sequences Mother﹠amp; PSNR value behind Daughter and the Highway Drive coding.Owing to do not have the compression ratio of equilibrium brightness, chromatic component between the JM conducting frame in the cataloged procedure, cause the PSNR value of entire image unbalanced.From table, can see, the high 0.4～1dB of the common specific luminance component of the PSNR of chromatic component, thereby wasted suitable code check, and to not too big improvement of picture quality." ACLE sequence PSNR " in table 1 and the table 2 is the data that adopt after chroma equalization method of the present invention is encoded.Can see that the PSNR value of chromatic component is near luminance component, each component equiblibrium mass distribution of entire image has been saved suitable code check, and by image observation being found the subjective quality difference of image is controlled at the degree that human eye is difficult to perceive.

Table 1 Mother﹠amp; Daughter

Table 2 Highway Drive

As shown in Figure 3 and Figure 4, Fig. 3 and Fig. 4 are the comparison diagrams as a result that the JM encoding software that adopts existing JM encoding software and add balance method of the present invention is encoded to standard test sequences Silent and Foreman respectively.Wherein, JM represents the test encoding software of JVT official issue; JM_ACLe represents JM is made amendment and adds the encoding software of equalization methods of the present invention.For the Silent sequence, under high code check situation, because image is relatively fine and smooth, space smoothing degree height, the effect that causes using behind the chroma equalization method of the present invention is preferable; Under the low code check situation, because picture noise strengthens, reference picture is spatially level and smooth inadequately, causes the effect of this method to descend.For the more violent Foreman sequence of motion, because the motion vector of 1/4 pixel precision increases, can see, the PSNR value of chromatic component is near luminance component, each component equiblibrium mass distribution of entire image, chroma equalization method of the present invention has been obtained better effect, can observe in the high code check part of Foreman, during identical PSNR value, code check has improved about 8%.

In the above-described embodiments, only the present invention has been carried out exemplary description, but those skilled in the art can design various execution modes according to different actual needs under the situation of the scope and spirit that do not break away from the present invention and protected.

Claims (4)

1. chroma equalization method that is used for video coding, described equalization methods comprises:

A. obtain the luminance component residual error SAD between present image and the predicted picture _YWith chromatic component residual error SAD _C

B. by described luminance component residual error SAD _YWith described chromatic component residual error SAD _CCalculate the residual error distribution SATD in the frequency domain _YAnd SATD _C

C. based on above-mentioned residual error distribution SATD _Y, SATD _CAnd the luminance quantization parameter QP that obtains in advance _YThe following formula of substitution draws chromaticity quantization parameter QP _C:

${\mathrm{Qp}}_C={\mathrm{Qp}}_Y{+\mathrm{\ΔQp}=\mathrm{Qp}}_Y+\mathrm{int}(\frac{1}{{\mathrm{Qp}}_Y/\mathrm{\ω}}+(\frac{{\mathrm{SATD}}_Y}{N_Y}-\frac{{\mathrm{SATD}}_C}{N_C})\·\mathrm{\σ});$

Wherein, int () represents bracket function, N _YExpression luminance pixel quantity, N _CExpression chroma pixel quantity, σ and ω are modifying factor, σ value 0.6, ω value 6,

D. respectively based on described luminance quantization parameter QP _YWith described chromaticity quantization parameter QP _CLuminance component and chromatic component to described present image are encoded.

2. equalization methods according to claim 1 is characterized in that: in described step b, and described residual error distribution SATD _Y, SATD _CBy the Hamilton conversion acquisition approximate with integer transform.

3. equalization methods according to claim 1 is characterized in that: the sample format of described present image and described predicted picture is YUV4: 2: 0.

4. equalization methods according to claim 1 is characterized in that: described chromatic component comprises U chromatic component or V chromatic component.

Images