CN1190969C - Method for controlling video coding bit rate - Google Patents

Abstract

The present invention discloses a method for controlling a video coding bit rate. The favorable linear relationship which exists between coding bit number and a conversion ratio parameter and the favorable information source weak correlation which exists between a coding distortion factor and the conversion ratio parameter are simultaneously utilized, so that the quick and accurate estimation or prediction of a coding bit rate is realized through the conversion ratio parameter; the bit number of goal coding is determined according to the predication result and the size of a coding buffer area; then, through the relation between the conversion ratio parameter and a quantization parameter, the coding quantization parameter is adjusted, and the control for the coding bit rate is realized; a coding parameter is adjusted according to a distortion factor of a coding macro-block. The present invention can realize the quick and accurate prediction of the coding bit rate, the quick and accurate optimization of coding bit allocation, and the quick control of the coding bit rate.

Description

A kind of video code bit rate control method

Technical field

The invention belongs to the transmission and the process field of the video information in the digital video communications field, be specifically related to a kind of video code bit rate control method.

Background technology

1948, Shannon proposed the coding thinking that interrelates with certain fidelity criterion.Deliver relevant paper in nineteen fifty-nine again afterwards, information source coding problem had been done the statement of essence, pointed out the theoretical boundary that rate distortion theory can the specified data compression, thus the basis of further having established rate distortion theory.After people's such as Jelinek, Gallager, Berger development has formed qualification distortion source encoding theory.

In limiting distortion source encoding theory, rate distortion (R-D) function R (D) is defined as: given information source message probability distribution { P (xi) } and average distortion function allow under the prerequisite of distortion D, these information source message of coding transmission, and when making its distortion level in allowed band, the minimum of needed information coding transmission coding bit rate.Based on the definition of R (D) function, can obtain to draw a conclusion: under the condition that allows certain distortion, coding bit rate can reduce, and needed channel capacity also can correspondingly reduce simultaneously.And the distortion factor that allows is big more, and then coding bit rate is can compressed degree also big more.On the other hand, from information source itself, if the distortion factor that allows is big more, then the incidence of information source (being the time entropy of information source) is can compressed degree also big more.This has just drawn the information source coding problem that limits distortion, also is the rationale of rate-distortion optimization in the digital signal encoding method.

In the digital video coding system, the coding bit rate of encoder is all relevant with quantization parameter with image fault, for the ease of analyzing, the general relation that adopts coding bit rate-quantization function R (q) to describe coding bit rate and quantization parameter adopts the relation between coding distortion-quantization function D (q) description image fault and the quantization parameter equally.By analysis, carry out the rate-distortion optimization of digital signal encoding, thereby realize coding bit rate control these two functions.Yet this coding bit rate function and distortion function based on quantization parameter all has very strong nonlinear characteristic, and very strong with the correlation of information source.Adopting this Mathematical Modeling is that coded system realizes coding bit rate-distortion analysis exactly, must set up very complicated coding bit rate function and distortion function.As U.S. Patent No. 6,366,704, No.6,356,668, No.6,212,302, No.6,037,987, No.6,072,831 and No.6,167, in 162, when solving coding bit rate control problem, all adopted coding bit rate-distortion model based on quantization parameter, these models or fairly simple, the second order polynomial approximate description non-uniform encoding bit rate function of employing quantization parameter makes its accuracy be subjected to certain limitation; Very complicated, adopt higher order polynomial to describe non-uniform encoding bit rate function, its accuracy is improved, but computation complexity is very high again, be difficult to be applied in the actual coded system.

Summary of the invention

The object of the present invention is to provide a kind of video code bit rate control method that can overcome above-mentioned defective, this method can realize coding bit rate control quickly and accurately.

For achieving the above object, a kind of video code bit rate control method is characterized in that comprising following step:

1. set up the mapping table between quantization parameter and the conversion ratio parameter, the corresponding relation between quantization parameter and the conversion ratio parameter passes through formula $p=1-\frac{1}{M}\underset{\left|x\right|<2q}{\mathrm{\&Sigma;}}{D}_0\left(x\right)-\frac{1}{M}\underset{\left|x\right|<2.5q}{\mathrm{\&Sigma;}}{D}_l\left(x\right)$ Obtain; Wherein, D ₀(X) and D _l(X) discrete cosine transform coefficient that is respectively intraframe coding and inter-coded macroblocks distributes, and x is the value of the conversion coefficient after quantizing, and q is a quantization parameter, and p is corresponding conversion ratio parameter, and M is the pixel count of present image;

2. determine the target code bit number of present image according to the size of encoding buffer;

3. utilize given conversion ratio parameter and quantize after the average information entropy of conversion coefficient between ratio, determine to reach the size of the corresponding conversion ratio parameter of target code bit number;

4. according to the mapping table between quantization parameter and the conversion ratio parameter, determine the quantization parameter of present image coding;

5. adjust the conversion ratio parameter of next macro block according to the distortion factor of current macro;

6. according to the mapping table between quantization parameter and the conversion ratio parameter, determine the quantization parameter of next macroblock coding; 5. repeating step finishes until all macroblock codings of a frame;

7. determine the target code bit number of next frame image according to the size of encoding buffer;

8. conversion ratio parameter value that is adopted according to present image coding and the ratio between the actual coding bit number

Be worth, determine the size of the conversion ratio parameter of next frame image encoding;

9. according to the mapping table between quantization parameter and the conversion ratio parameter, determine the next frame image encoding

Quantization parameter;

10. repeating step 5.-9., up to end-of-encode.

The present invention is used for the digital video coding system, on the basis of the rate-distortion model that encoding bit rate control method adopted of traditional video coding based on quantization parameter, the new rate-distortion model that adopted the neotectonics of conversion ratio parameter renegotiation, and the corresponding relation by quantization parameter and conversion ratio parameter, realize coding bit rate control quickly and accurately.The present invention has following major advantage:

One. have good linear relationship between conversion ratio parameter and the number of coded bits, help realizing coding bit rate prediction quickly and accurately.

They are two years old. and have the weak correlation of good information source between conversion ratio parameter and the coding distortion degree, help realizing coded-bit allocation optimized quickly and accurately.

They are three years old. and there is corresponding relation between conversion ratio parameter and the quantization parameter,, can realizes the control of coding bit rate fast by the mode of look-up table according to this corresponding relation.

Description of drawings

The flow chart of Fig. 1 encoding bit rate control method of the present invention.

Embodiment

Video code bit rate control method provided by the invention is mainly used in the digital video coding system of solution based on conversion, realizes coding bit rate control quickly and accurately.Specific implementation method is finished according to the following steps.

The first step: set up the mapping table between quantization parameter and the conversion ratio parameter.

The conversion ratio parameter p is: original image has the number N of the coefficient (not being quantified as zero coefficient) of coding practical significance through after the discrete cosine transform _TWith the original pixels number N ₀Ratio, i.e. p=N _T/ N ₀Corresponding relation between quantization parameter and the conversion ratio parameter passes through formula $p=1-\frac{1}{M}\underset{\left|x\right|<2q}{\mathrm{\&Sigma;}}{D}_0\left(x\right)-\frac{1}{M}\underset{\left|x\right|<2.5q}{\mathrm{\&Sigma;}}{D}_l\left(x\right)$ Obtain.Wherein, D ₀(X) and D _l(X) discrete cosine transform coefficient that is respectively intraframe coding and inter-coded macroblocks distributes, and x is the value of the conversion coefficient after quantizing, and q is a quantization parameter, and p is corresponding conversion ratio parameter, and M is the pixel count of present image.

Second step: the target code bit number of determining present image according to the size of encoding buffer.

Adopt formula $R_T=\frac{C}{F}-(B-0.2\&times;B_T)-R_{\mathrm{MV}}-R_H$ Obtain the target code bit number of present image.Wherein, it is channel width that B has used buffer size, C for coding, and F is a frame per second, B _TBe buffer sizes, R _MVAnd R _HBe respectively the number of coded bits of motion vector and header.Also can adopt formula $R_T=\frac{C}{F}\&times;\frac{2.4\&times;B_T-B}{2\&times;B_T+B}-R_{\mathrm{MV}}-R_H$ Obtain the target code bit number of present image.Its objective is according to the full state of buffer empty and adjust present image preassignment number of coded bits.This step 1 aspect is the control channel transmission delay effectively, can adjust each frame preassignment number of coded bits in conjunction with channel actual speed rate state on the other hand, can effectively utilize the channel current bandwidth.As B=0.2 * B _T, promptly the encoding buffer fullness degree is 20% o'clock of total buffer size, present frame preassignment bit number R _TEqual average number of coded bits $\frac{C}{F}-R_{\mathrm{MV}}-R_H;$ B＞0.2 * B _TThe time, preceding frame preassignment bit number R _TLess than average number of coded bits $\frac{C}{F}-R_{\mathrm{MV}}-R_H,$ To prevent the encoding buffer overflow, can effectively control coding transmission and postpone; B＜0.2 * B _TThe time, preceding frame preassignment bit number R _TGreater than average number of coded bits $\frac{C}{F}-R_{\mathrm{MV}}-R_H,$ To prevent the encoding buffer underflow, can effectively utilize current transmission bandwidth;

The 3rd step: utilize the ratio between the average information entropy of conversion coefficient after given conversion ratio parameter and the quantification, determine to reach the size of the corresponding conversion ratio parameter of target code bit number.

Pass through formula $p_T=\frac{R_T}{H_0}\&CenterDot;p_0$ Determine the size of the pairing conversion ratio parameter of target code bit number, wherein p _TBe the size of the pairing conversion ratio parameter of target code bit number, R _TBe target code bit number, p ₀For quantization parameter is made as the size of 16 o'clock pairing conversion ratio parameters, H ₀For quantization parameter is made as 16 o'clock average information entropys through the conversion coefficient after quantizing.

The 4th step:, determine the quantization parameter of present image coding according to the mapping table between quantization parameter and the conversion ratio parameter.

The 5th step: the conversion ratio parameter of adjusting next macro block according to the distortion factor of current macro.

For first macro block of current encoded image, adopt the conversion ratio parameter of present image to encode.For the conversion ratio parameter of next coded macroblocks, by formula $p_i=\frac{1}{a_i}[\ln \frac{a_i{\mathrm{\&sigma;}}_i^2}{{\mathrm{\&theta;}}_i}+\frac{1}{{\mathrm{\&Sigma;}}_{i=1}^L{\mathrm{\&theta;}}_i}(R_T-{\mathrm{\&Sigma;}}_{i=1}^L\frac{{\mathrm{\&theta;}}_i}{a_i}\&CenterDot;\ln \frac{a_i{\mathrm{\&sigma;}}_i^2}{{\mathrm{\&theta;}}_i})]$ Obtain.Wherein, θ _i=R _i/ p _i, R _iBe the actual coding bit number of current macro, p _iBe the conversion ratio parameter of current macro, $a_i=\frac{1}{p_i}\ln \frac{{\mathrm{\&sigma;}}_i}{D_i},$ σ _iBe the standard deviation of current macro, D _iBe the distortion factor of current macro, R _TBe the target code bit number, the macroblock number of L for having encoded.

The 6th step:, determine the quantization parameter of next macroblock coding according to the mapping table between quantization parameter and the conversion ratio parameter.

The 7th step: the target code bit number of determining the next frame image according to the size of encoding buffer.Specific implementation method is identical with second step.

The 8th step: conversion ratio parameter value that is adopted according to present image coding and the ratio between the actual coding bit number, determine the size of the conversion ratio parameter of next frame image encoding.Pass through formula $p_{n+1}=\frac{R_T}{R_n}\&CenterDot;p_n$ Determine the conversion ratio parameter of next frame image encoding, wherein P _N+lBe the conversion ratio parameter of next frame image encoding, R _TBe target code bit number, P _nBe the conversion ratio parameter that the present image coding is adopted, R _nActual bit number for the present image coding.

The 9th step:, determine the quantization parameter of next frame image encoding according to the mapping table between quantization parameter and the conversion ratio parameter;

The tenth step repeated the 5th and went on foot for the 9th step, up to end-of-encode.

Can see that the operand of above method mainly concentrates in the distributed process that obtains conversion coefficient, the calculating in other link has only addition and multiplying.With respect to whole encoding process process, computation complexity is very low.

The bit-rate control method that this patent proposes can be used for the video coding system of various criterion.Here be example with MPEG-2 TM5 coded system, the effect of this video code bit rate control method is described.In MPEG-2 TM5 coded system, the video code bit rate control method that adopts this patent to propose substitutes original method, and cycle tests " Foreman ", " Salesman ", " Tabletennis " and " Coastguard " are encoded respectively.

In order to describe the execution efficient of coding bit rate control, definition control corresponding error is:

$E_{\mathrm{rc}}=\frac{R-R_T}{R_T}\&times;100\%$

Wherein R and R _TReality and the target code bit number of representing every two field picture respectively.Listed experimental result in the table 1 respectively.As can be seen, the video code bit rate control method that this patent proposes has effectively realized the control of coding bit rate under the condition of given target code bit rate, and departure has been compared obvious reduction with original MPEG-2TM5 system, is less than about 2% greatly.Simultaneously, picture quality also is significantly improved.Mean P SNR has improved 0.87dB.

Table 1 is used for this method the coding bit rate control result of MPEG-2 TM5

Video sequence		Foreman	Salesman	Tabletennis	Costguard	Coastguard
							Target bit rate		288000	288000	288000	288000	1024000
The actual coding bit rate	The former method of TM5	289024	289089	288208	288895	1024273
							This method	288203	288170	288008	288169	1024152
	Departure	The former method of TM5	12.8％	4.9％	15.2％	6.5％							2.9％
This method							0.9％	1.3％	0.7％	0.8％	0.3％
		PSNR(dB)	The former method of TM5	32.96	36.58	32.24						31.38	38.37
This method	33.84						37.20	33.12	32.20	39.64
			Improve	0.88	0.62	0.88					0.82	1.17

Claims (6)

1. video code bit rate control method is characterized in that comprising following step:

1. set up the mapping table between quantization parameter and the conversion ratio parameter, the corresponding relation between quantization parameter and the conversion ratio parameter passes through formula $p=1-\frac{1}{M}\underset{\left|x\right|<2q}{\mathrm{\&Sigma;}}{D}_0\left(x\right)-\frac{1}{M}\underset{\left|x\right|<2.5q}{\mathrm{\&Sigma;}}{D}_1\left(x\right)$ Obtain; Wherein, D ₀(x) and D ₁(x) discrete cosine transform coefficient that is respectively intraframe coding and inter-coded macroblocks distributes, and x is the value of the conversion coefficient after quantizing, and q is a quantization parameter, and p is corresponding conversion ratio parameter, and M is the pixel count of present image;

2. determine the target code bit number of present image according to the size of encoding buffer;

3. utilize given conversion ratio parameter and quantize after the average information entropy of conversion coefficient between ratio, determine to reach the size of the corresponding conversion ratio parameter of target code bit number;

4. according to the mapping table between quantization parameter and the conversion ratio parameter, determine the quantization parameter of present image coding;

5. adjust the conversion ratio parameter of next macro block according to the distortion factor of current macro;

6. according to the mapping table between quantization parameter and the conversion ratio parameter, determine the quantization parameter of next macroblock coding; 5. repeating step finishes until all macroblock codings of a frame;

7. determine the target code bit number of next frame image according to the size of encoding buffer;

8. conversion ratio parameter value that is adopted according to present image coding and the ratio between the actual coding bit number are determined the size of the conversion ratio parameter of next frame image encoding;

9. according to the mapping table between quantization parameter and the conversion ratio parameter, determine the quantization parameter of next frame image encoding;

10. repeating step 5.-9., up to end-of-encode.

2. video code bit rate control method according to claim 1 is characterized in that, described step 2. with step 7. in, adopt formula $R_T=\frac{C}{F}-(B-0.2\&times;B_T)-R_{\mathrm{MV}}-R_H$ Obtain the target code bit number of present image; Wherein, it is channel width that B has used buffer size, C for coding, and F is a frame per second, B _TBe buffer sizes, B _MVAnd R _HBe respectively the number of coded bits of motion vector and header.

3. video code bit rate control method according to claim 1 is characterized in that, described step 2. with step 7. in, adopt formula $R_T=\frac{C}{F}\&times;\frac{2.4\&times;B_T-B}{2\&times;B_T+B}-R_{\mathrm{MV}}-R_H$ Obtain the target code bit number of present image, wherein, it is channel width that B has used buffer size, C for coding, and F is a frame per second, B _TBe buffer sizes, R _MVAnd R _HBe respectively the number of coded bits of motion vector and header.

4 according to claim 1,2 or 3 described video code bit rate control methods, it is characterized in that, described step 3. in, pass through formula $p_T=\frac{R_T}{H_0}\&CenterDot;p_0$ Determine the size of the pairing conversion ratio parameter of target code bit number, wherein p _TBe the size of the pairing conversion ratio parameter of target code bit number, R _TBe target code bit number, p ₀For quantization parameter is made as the size of 16 o'clock pairing conversion ratio parameters, H ₀For quantization parameter is made as 16 o'clock average information entropys through the conversion coefficient after quantizing.

5 video code bit rate control methods according to claim 4 is characterized in that, described step 5. in, for first macro block of current encoded image, adopt the conversion ratio parameter of present image to encode; For the conversion ratio parameter of next coded macroblocks, by formula $p_i=\frac{1}{a_i}[\ln \frac{a_i{\mathrm{\&sigma;}}_i^2}{{\mathrm{\&theta;}}_i}+\frac{1}{{\mathrm{\&Sigma;}}_{i=1}^L{\mathrm{\&theta;}}_i}(R_T-{\mathrm{\&Sigma;}}_{i=1}^L\frac{{\mathrm{\&theta;}}_i}{a_i}\&CenterDot;\ln \frac{a_i{\mathrm{\&sigma;}}_i^2}{{\mathrm{\&theta;}}_i})]$ Obtain; Wherein, R _TBe the target code bit number, the macroblock number of L for having encoded, θ _i=R _i/ p _i, R _iBe the actual coding bit number of current macro, p _iBe the conversion ratio parameter of current macro, $a_i=\frac{1}{p_i}\ln \frac{{\mathrm{\&sigma;}}_i}{D_i},$ σ _iBe the standard deviation of current macro, D _iThe distortion factor for current macro.

6 video code bit rate control methods according to claim 5 is characterized in that, described step 8. in, pass through formula $p_{n+1}=\frac{R_T}{R_n}\&CenterDot;p_n$ Determine the conversion ratio parameter of next frame image encoding, wherein p _N+1Be the conversion ratio parameter of next frame image encoding, R _TBe target code bit number, p _nBe the conversion ratio parameter that the present image coding is adopted, R _nActual bit number for the present image coding.

Images