CN100581262C

CN100581262C - Code rate controlling method for video coding based on Rho domain

Info

Publication number: CN100581262C
Application number: CN 200810112629
Authority: CN
Inventors: 戴琼海; 肖红江; 陆峰
Original assignee: Tsinghua University
Current assignee: Tsinghua University
Priority date: 2008-05-23
Filing date: 2008-05-23
Publication date: 2010-01-13
Anticipated expiration: 2028-05-23
Also published as: CN101287123A

Abstract

The invention relates to a video coding code rate control method based on a Rho domain, which pertains to the technical field of multimedia communication; the invention comprises the steps as follows: the bit rate of a group of pictures (GOP) is confirmed according to a given target code rate, and the bit rate distribution of frame-level is further respectively confirmed according to three frames which are an I-frame (Intra-frame frame), a P-frame (unidirectional predictive frame) and a B-frame (bidirectional predictive frame); the characteristic parameters of the previous frames of the same type are checked to predict the slope parameter Theta of a linear code rate model, model interception difference Delta c, the ratio s of motion vector code rate and entropy coding code rate that is quantified by residual coefficient, and zero coefficient proportion Rho that meets the code rate limit is calculated; a Rho-QP mapping table is referred to obtain a quantization parameter (QP) which is used for coding the current video frame. The code rate control method of the invention is simple and practical, and has excellent performance and the original video coding standards.

Description

Video coding rate control method based on rho domain

Technical Field

The invention belongs to the technical field of multimedia communication, and particularly relates to a low-complexity video coding rate control method based on a rho domain.

Background

In video communication, network bandwidth for carrying data streams is time-varying and limited, and factors to be considered by researchers are not only pure compression efficiency, but also the matching relationship between channel bandwidth and information source rate, that is, rate control is performed on video coding to fully utilize channels. For the existing block-based hybrid video coding scheme, such as MPEG-4, the existing code rate control method usually starts from the aspects of frame type (I frame, P frame and B frame), group of picture (GOP) structure (IPP, IBP, IBBP, etc.), buffer limitation, bit rate model, computational complexity, etc., and approaches the expected bit number at the frame level or macroblock level by dynamically adjusting the encoder parameters such as Quantization Parameter (QP). However, the newly established video coding standard h.264 introduces many new features. On one hand, 21 prediction modes including intra-frame, inter-frame and SKIP (SKIP) greatly reduce the precision of a code rate control algorithm aiming at the conventional video coding standard, and even fail. On the other hand, Rate Distortion Optimization (RDO) couples quantization parameters that are not originally related to motion vectors, and global rate control becomes more difficult. More importantly, in h.264, QP is used as an input of RDO process, and then the optimal QP is estimated in reverse according to the output of RDO, and the computation overhead of such fully-coupled iterative optimization is not tolerable in real-time encoding.

He et al, in an article "unified rate-distortion analysis framework in transform coding" (a unified rate-distortion analysis framework for transform coding) published in the international society of electrical and electronics engineers (ieee transactions, on Circuits and Systems for Video Technology) journal of the society of electrical and electronics engineers, states that: in video coding, each quantization parameter QP corresponds to a fixed quantization step q, and there is a rough one-to-one mapping relationship between the quantization step q and the zero coefficient ratio ρ. Then, for the transform-coded residual coefficients (I frame is intra prediction residual, P and B frames are motion compensation residual), the mapping of this zero coefficient ratio ρ to the argument q can be obtained from the distribution of statistical transform coefficients, i.e.:

<math> <mrow> <mi>ρ</mi> <mrow> <mo>(</mo> <mi>q</mi> <mo>)</mo> </mrow> <mo>=</mo> <mfrac> <mn>1</mn> <mi>L</mi> </mfrac> <msubsup> <mo>&Integral;</mo> <mrow> <mo>-</mo> <mn>2</mn> <mi>q</mi> </mrow> <mrow> <mo>+</mo> <mn>2</mn> <mi>q</mi> </mrow> </msubsup> <msub> <mi>D</mi> <mn>0</mn> </msub> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mi>dx</mi> <mo>+</mo> <mfrac> <mn>1</mn> <mi>L</mi> </mfrac> <msubsup> <mo>&Integral;</mo> <mrow> <mo>-</mo> <mn>2.5</mn> <mi>q</mi> </mrow> <mrow> <mo>+</mo> <mn>2.5</mn> <mi>q</mi> </mrow> </msubsup> <msub> <mi>D</mi> <mn>1</mn> </msub> <mrow> <mo>(</mo> <mi>x</mi> <mo>)</mo> </mrow> <mi>dx</mi> </mrow> </math>

wherein L is the coefficient number of the current video frame, D₀(x) And D₁(x) Is a statistical histogram (including ac and dc coefficients) of the residual coefficients after discrete cosine transform. Thus, the p-QP mapping table is obtained for the link with the quantization step q. Meanwhile, z.he also proposes a linear source-rate model in this article:

R(ρ)＝θ·(1-ρ)

and an analysis based on the zero coefficient ratio is called a ρ -domain analysis. Where θ is the slope parameter, ρ is the proportion of the quantized zero coefficient, and R is the residual coefficient coding bit rate in bits per pixel (bpp). The disadvantage of this model is that the accuracy test of this model is only applicable to the international image coding standard JEPG, the international video coding standards MPEG-2, h.263 and MPEG-4, but not very applicable to the video coding standard h.264 (more SKIP mode macro blocks, 4x4 integer transform and scale transform, etc.) with a relatively large zero coefficient ratio, and further modification is needed.

In addition, for the low complexity video coding rate control method, the following patents can be found at present:

(a) the patent with application number 200610052814.5 discloses a video compression code rate control method based on a low memory consumption lookup table;

(b) the patent with application number 200510073985.1 discloses a low-complexity integral code rate control method;

(c) the patent with application number 200510135494.5 discloses a code rate control method based on the difference histogram statistics of the local motion of a video sequence;

although the code rate control methods ensure low complexity, the extracted video content feature description mode is not better than rho domain characteristics and can reflect the features of the video content, and has certain influence on the precision of a code rate control algorithm, so that a general control method which has low calculation complexity and high code rate control precision is required to be found.

Disclosure of Invention

The invention aims to overcome the defects of the prior art, and provides a low-complexity video coding rate control method based on rho domain based on the linear characteristic of spatial domain and time domain continuity of video content, which has the characteristics of simplicity, practicability, superior performance, and extremely low calculation complexity and memory requirement, and is not only suitable for the former video coding standards H.261, H.263, MPEG-2 and MPEG-4, but also suitable for the latest H.264 standard.

The invention provides a video coding rate control method based on rho domain, which is characterized in that a given target code rate

The coding end performs a GOP-level and frame-level bit rate (bit/pixel unit) allocation algorithm based on a rho domain model to realize code rate control, and the coding end specifically comprises the following steps:

1) precoding a group of pictures to obtain initial characteristic parameters of a code rate control method

A group of GOP frames of a video sequence are coded by an initial quantization parameter QP, and a characteristic parameter { delta ] of each frame is recorded in an information base_i，ρ_i，θ_i，Δc_i，s_iIn which the letter subscript i ∈ c[1，L_GOP]∩Z⁺Representing the code frame number in the GOP, and taking a positive integer; delta_i，ρ_i，θ_i，Δc_iResidual texture complexity parameters, zero coefficient proportion parameters, code rate model slope parameters and code rate model data interception difference parameters in a rho domain model corresponding to the ith frame; s_iEncoding bit rate R for motion vectors in ith frame_j ^mvAnd entropy coding bit rate R after residual coefficient quantization_j ^coeThe ratio of (A) to (B);

2) calculating a GOP bit rate budget based on a target code rate

(21) Calculating the bit rate allocable for a GOP

Updating remaining available bit rate(first run, right of equation

Initialized to zero); wherein L is_GOPGOP length (in frames), f video coding frame rate (in frames/second);

(22) if the video frame type is I frame (intra frame), skipping to step 3);

if the frame type is a P frame (unidirectional predicted frame) or a B frame (bidirectional predicted frame), it jumps to step 4).

3) According to the remaining available bit rate

Calculating an expected I-frame bit rate

And the corresponding zero coefficient ratio rho, skipping to step 5);

4) according to the residueAvailable bit rate

Calculating a desired P-frame or B-frame bit rate and a corresponding zero coefficient ratio ρ;

5) estimating quantization coefficients QP of the current frame according to the desired zero coefficient ratio rho_j

(51) Inquiring a rho-QP mapping table to obtain conversion from a zero coefficient ratio rho to a quantization parameter QP;

(52) detecting the variation amplitude of the quantization parameter, and limiting the variation amplitude within delta QP;

6) with quantization parameter QP_jEncoding a current video frame

(61) Encoding a video frame to obtain a true output bit rate for the frame

(62) Calculating a remaining available bit rate

Wherein the symbol max {. cndot } represents the maximum of the two numbers;

7) recording the actual coding characteristic parameter of the current frame in an information base, and selecting a jump position according to a frame number:

(71) storing the characteristic parameters { delta, rho, theta, deltac, s } of the current frame in an information base (for the characteristic parameter prediction of the subsequent video frame);

(72) if frame number j ≠ L_GOPIf j is j +1, go to step 4);

(73) if the frame number j is L_GOPThen let j equal 1 and jump to step 2).

The invention is simple and practical, has excellent performance and mainly has the following beneficial effects:

(a) a general processing framework is provided for video rate control based on transform coding, wherein the form of a modified rho domain bit rate model is more general, and is suitable for not only previous video coding standards H.261, H.263, MPEG-2 and MPEG-4, but also the latest H.264 standard;

(b) the method has extremely low computational complexity and memory requirements, only linear regression is carried out on parameter sample values of a plurality of historical airspace linear characteristics rho domain models, and then the coding rate of the current frame can be predicted by means of rho variables;

(c) the prediction process is error-free and can be dynamically adjusted adaptively to follow the input video data.

Drawings

Fig. 1 is a flow chart of a rho domain-based video coding rate control method according to the present invention.

Detailed Description

The video coding rate control method based on rho domain provided by the invention is explained in detail by combining the attached drawings and the embodiment as follows:

in the present invention:

(a) the spatial domain linearity is characterized by the p-domain model (a modified model for h.264 for the linear source-rate model for z.he in the introduction of the background art),

R(ρ)＝θ·(1-ρ)+Δc

wherein rho is the proportion of the quantized zero coefficient in the current frame, deltac is the intercept difference, and R (rho) is the coding rate of the video single frame under the condition of the zero coefficient proportion rho; the model slope parameter θ is related to the video texture complexity δ of the current frame, and is further defined by a slope-texture complexity model

θ＝σ²·e^α(1-δ)

Where δ is measured by the normalized mean absolute error (MAD), which is the sum of the absolute values of all residual coefficients between the current frame and the predicted frame divided by 255. σ and α are model parameters.

(b) Temporal continuity is characterized by the parameters { θ, Δ c, s } of previously encoded frames; wherein, theta and delta c are parameters in a space domain model, and s is the ratio of the motion vector code rate in a frame to the entropy coding code rate after residual error coefficient quantization;

the method according to the invention is characterized in that a target code rate is specifiedThe coding end performs a GOP-level and frame-level bit rate (unit is bit/pixel) allocation algorithm based on a rho-domain model to realize code rate control, and the flow of the method is shown in fig. 1, and specifically comprises the following steps:

A group of GOP frames of a video sequence are coded by an initial quantization parameter QP, and a characteristic parameter { delta ] of each frame is recorded in an information base_i，ρ_i，θ_i，Δc_i，s_iIn which the letter subscript i ∈ [1, L ]_GOP]∩Z⁺Representing the code frame number in the GOP, and taking a positive integer; delta_i，ρ_i，θ_i，Δc_iResidual texture complexity parameters, zero coefficient proportion parameters, code rate model slope parameters and code rate model data interception difference parameters in a rho domain model corresponding to the ith frame; s_iEncoding bit rate R for motion vectors in ith frame_j ^mvAnd entropy coding bit rate R after residual coefficient quantization_j ^coeThe ratio of (A) to (B);

2) calculating a GOP bit rate budget based on a target code rate

(21) Calculating the bit rate allocable for a GOPUpdating remaining available bit rate

(first run, right of equationInitialized to zero); wherein L is_GOPGOP length (in frames), f video coding frame rate (in frames/second);

(22) if the video frame type is I frame (intra frame), skipping to step 3);

3) According to the remaining available bit rate

Calculating an expected I-frame bit rateAnd the corresponding zero coefficient ratio rho, skipping to step 5);

4) according to the remaining available bit rate

6) with quantization parameter QP_jEncoding a current video frame

(61) Encoding a video frame to obtain a true output bit rate for the frame

(62) Calculating a remaining available bit rate

Wherein the symbol max {. cndot } represents the maximum of the two numbers;

(72) if frame number j ≠ L_GOPIf j is j +1, go to step 4);

(73) if the frame number j is L_GOPThen let j equal 1 and jump to step 2).

According to the residual available bit rate in the step 3) above

Calculating an expected I-frame bit rate

And a corresponding zero coefficient ratio p, comprising the steps of:

(31) calculating an expected I-frame bit rate

Wherein, w_I，w_P，w_BDesired bit rate weights, 1, y, for single frame video of type I, P, B, respectively_P，γ_BThe number of GOPs for which they occupy;

(32) updating the characteristic parameters { slope theta and intercept difference delta c } of the current I frame, namely using the characteristic parameters { slope theta and intercept difference delta c } of the previous I frame adjacent to the current I frame as the characteristic parameters { slope theta and intercept difference delta c } of the current frame;

(33) calculating an expected zero coefficient ratio

Jumping to step 5); wherein,is the bit rate of the video frame header;

according to the residual available bit rate in the step 4) above

Calculating the expected bit rate of the P frame or the B frame and the corresponding zero coefficient proportion rho, and specifically comprising the following steps:

(41) calculating an available bit rate of the current frame j;

(411) if the available bit rate remains

Then frame skipping is performed (without encoding the frame);

(412) if the available bit rate remains

Then

(a) If the current frame is a P frame, the frame is assigned a desired bit rate of:

the number of P frames in the GOP waiting for allocated bit rate minus 1:

γ_P＝γ_P-1；

(b) if the current frame is a B frame, the frame is assigned a desired bit rate of:

the number of B frames in the GOP waiting for allocated bit rate minus 1:

γ_B＝γ_B-1；

(42) detecting characteristic parameters of the previous similar type frame, and updating the characteristic parameters { slope theta, intercept difference delta c and bit rate ratio s } of the current frame;

(43) calculating the desired bit rate of the current frame

Limiting the corresponding zero coefficient ratio:

the step 42) of detecting the characteristic parameters of the previous frame of the same type, and updating the characteristic parameters { slope θ, intercept difference Δ c, and bit rate ratio s } of the current frame specifically includes the following steps:

(421) taking N nearest video frames of the same type before the current frame as prediction reference frames, and taking N groups of characteristic parameter values { theta ] corresponding to the N nearest video frames of the same type from an information base_i，δ_i，Δc_i，s _i1 ≦ i ≦ N, and is converted to { ln θ ≦ N_i，δ_i，Δc_i，s_i}; wherein N is a positive integer, and the subscript i belongs to Z⁺Indicating adjacent reference frame numbers; theta_iLinear rate model slope, delta, for the ith reference frame_iResidual texture complexity for the ith reference frame;

(422) linearizing a slope-texture complexity model to

lnθ(δ)＝2lnσ+α(1-δ)

And performing least square fitting on the N groups of sample point values, solving model parameters alpha and sigma, and updating characteristic parameters { theta, delta c, s }:

(a) estimating slope-texture complexity model parameter values for a current encoded frame

<math> <mrow> <mi>α</mi> <mo>=</mo> <mfrac> <mrow> <mover> <mi>δ</mi> <mo>&OverBar;</mo> </mover> <mrow> <mo>(</mo> <mi>ln</mi> <munderover> <mi>Π</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msub> <mi>θ</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> <mo>-</mo> <mrow> <mo>(</mo> <munderover> <mi>Σ</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msub> <mi>δ</mi> <mi>i</mi> </msub> <mi>ln</mi> <msub> <mi>θ</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> </mrow> <mrow> <munderover> <mi>Σ</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msubsup> <mi>δ</mi> <mi>i</mi> <mn>2</mn> </msubsup> <mo>-</mo> <mi>n</mi> <msup> <mover> <mi>δ</mi> <mo>&OverBar;</mo> </mover> <mn>2</mn> </msup> </mrow> </mfrac> <mo>,</mo> </mrow> </math>

<math> <mrow> <mi>σ</mi> <mo>=</mo> <mi>exp</mi> <mo>[</mo> <mfrac> <mrow> <mfrac> <mn>1</mn> <mi>n</mi> </mfrac> <mrow> <mo>(</mo> <mi>ln</mi> <munderover> <mi>Π</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msub> <mi>θ</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> <mrow> <mo>(</mo> <munderover> <mi>Σ</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msubsup> <mi>δ</mi> <mi>i</mi> <mn>2</mn> </msubsup> <mo>)</mo> </mrow> <mo>-</mo> <mrow> <mo>(</mo> <mover> <mi>δ</mi> <mo>&OverBar;</mo> </mover> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> <mrow> <mo>(</mo> <munderover> <mi>Σ</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msub> <mi>δ</mi> <mi>i</mi> </msub> <mi>ln</mi> <msub> <mi>θ</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> <mo>+</mo> <mi>n</mi> <mover> <mi>δ</mi> <mo>&OverBar;</mo> </mover> <mrow> <mo>(</mo> <mi>ln</mi> <munderover> <mi>Π</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msub> <mi>θ</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> </mrow> <mrow> <mn>2</mn> <mrow> <mo>(</mo> <munderover> <mi>Σ</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msubsup> <mi>δ</mi> <mi>i</mi> <mn>2</mn> </msubsup> <mo>-</mo> <mi>n</mi> <msup> <mover> <mi>δ</mi> <mo>&OverBar;</mo> </mover> <mn>2</mn> </msup> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>]</mo> </mrow> </math>

Wherein,

<math> <mrow> <mover> <mi>δ</mi> <mo>&OverBar;</mo> </mover> <mo>=</mo> <mfrac> <mrow> <munderover> <mi>Σ</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <msub> <mi>δ</mi> <mi>i</mi> </msub> </mrow> <mi>N</mi> </mfrac> <mo>;</mo> </mrow> </math>

(b) updating the characteristic parameters { theta, deltac, s } of the current frame:

rho-domain bit rate model slope value theta-sigma²e^α(1-δ)；

Rho-domain bit rate model truncation data difference

Ratio of motion vector coding bit rate to entropy coding bit rate after residual coefficient quantization

The method of the invention is further described in detail with reference to the accompanying drawings and examples.

The conditions of this example are as follows:

setting a reference software JM of an international video coding standard H.264 adopted by an encoder; an encoder GOP structure is ibpbp. I. P, B frame type single frame expected bit number weight is w_I＝10，w_P1.5 and w _B1, the number of which in a GOP is 1, 7 and 7 respectively; the coding frame rate f is 30. The test sequence adopts a Foreman sequence of a standardized image format (CIF, 352x 288);

the method of the embodiment comprises the following steps: given target code rate

(unit is bit/pixel/second), the encoder can perform a GOP level and frame level bit rate (unit is bit/pixel) allocation algorithm based on the rho domain model to realize rate control, and the method specifically comprises the following steps:

A GOP of a video sequence is encoded with an initial quantization parameter QP 28, and a characteristic parameter { δ } of each frame is recorded in an information base_i，ρ_i，θ_i，Δc_i，s_iIn which the letter subscript i ∈ [1, L ]_GOP]∩Z⁺Indicating the code frame number, Z, in the present GOP⁺Represents taking a positive integer; delta_i，ρ_i，θ_i，Δc_iResidual texture complexity parameters, zero coefficient proportion parameters, code rate model slope parameters and code rate model data interception difference parameters in a rho domain model corresponding to the ith frame; s_iEncoding bit rate R for motion vectors in ith frame_j ^mvAnd entropy coding bit rate R after residual coefficient quantization_j ^coeThe ratio of (A) to (B);

2) calculating a GOP bit rate budget based on a target code rate

(21) Calculate a GOP energyAllocated bit rateUpdating remaining available bit rate

(first run, right of equation

Initialized to zero);

(22) if the video frame type is I frame, skipping to step 3);

if the frame type is P or B frame, jump to step 4).

3) According to the remaining available bit rateCalculating an expected I-frame bit rate

And a corresponding zero coefficient ratio p, comprising the steps of:

(31) calculating an expected I-frame bit rate

(33) calculating an expected zero coefficient ratio

Jumping to step 5); wherein,

is the bit rate of the video frame header;

4) according to the remaining available bit rate

Calculating a desired P-frame or B-frame bit rate and a corresponding zero coefficient ratio ρ, comprising the steps of:

(41) calculating an available bit rate of the current frame j;

(411) if the available bit rate remains

Then frame skipping is performed (without encoding the frame);

(412) if the available bit rate remains

Then

the number of P frames in the GOP waiting for allocated bit rate is reduced by one:

γ_P＝γ_P-1；

the number of B frames in the GOP waiting for allocated bit rate is reduced by one:

γ_B＝γ_B-1；

(42) detecting the characteristic parameters of the previous frames of the same type, and updating the characteristic parameters { slope theta, intercept difference delta c and bit rate ratio s } of the current frame, namely predicting by the characteristic parameters of the two frames of the same type which are most adjacent to the current frame (the P frame is predicted by the two previous P frames, and the B frame is predicted by the two previous B frames).

(421) Extracting 2 groups of model parameter values (theta) corresponding to 2 nearest video frames of the same type before the current frame from the information base₁，δ₁，Δc₁，s₁}，{θ₂，δ₂，Δc₂，s₂And is converted into { ln θ }₁，δ₁，Δc₁，s₁}，{lnθ₂，δ₂，Δc₂，s₂}; wherein, theta₁，θ₂The slope of the bit rate model, δ, for the 1 st and 2 nd reference frames, respectively₁，δ₂Residual texture complexity, Δ c, for 1, 2 reference frames₁，Δc₂Is the intercept data difference of the linear code rate model of the 1 st and 2 nd reference frames, s₁，s₂Is the ratio of the motion vector code rate of the 1 st and 2 nd reference frames and the entropy coding code rate after residual error coefficient quantization.

(422) Linearization with slope-texture complexity model

lnθ(δ)＝2lnσ+α(1-δ)

And performing least square fitting on the 2 groups of sample values, solving model parameters alpha and sigma, and updating characteristic parameters { theta, delta c, s }:

rho domain code rate model slope value theta-sigma²e^α(1-δ)；

Rho domain code rate model intercept data difference

Ratio of motion vector code rate to entropy coding code rate after residual error coefficient quantization

s = \frac{s_{1} + s_{2}}{2};

(43) Calculating the desired bit rate of the current frame

Limiting the corresponding zero coefficient ratio:

(51) Inquiring a rho-QP mapping table to obtain conversion from a zero coefficient ratio rho to a quantization parameter QP; wherein the p-QP mapping table is obtained in the manner introduced in the background art.

In this embodiment, the ρ -QP mapping table is exemplified by a Foreman standard test sequence with medium motion complexity in CIF format, and for an I frame, the mapping table is as follows:

for P and B frames, the mapping tables are as follows:

that is, the QP value is set when the calculated ρ is closest to the boundary of the section. For example, for an I frame, if ρ is 0.953, it belongs to the interval [0.951, 0.95793], nearest to 0.951, so the quantization parameter 33 corresponding to 0.951 is taken;

(52) detecting the variation amplitude of the quantization parameter, and limiting the variation amplitude to be within the range of delta QP +/-2; that is, if the last frame QP was 26, but the QP given in step (51) was 29 (or 23), then at most only QP 28 (or 24) could be obtained; if the QP given in step (51) is 27, it is in the range of [24, 28], so the current frame does not have to be clipped.

6) With quantization parameter QP_jEncoding a current video frame

(61) Encoding a video frame to obtain a true output bit rate for the frame

(62) Calculating a remaining available bit rate

Wherein the symbol max {. cndot } represents the maximum of the two numbers;

7) recording the actual coding characteristic parameter of the current frame in the information base, and selecting the jump position according to the frame number

(72) if frame number j ≠ L_GOPIf j is j +1, go to step 4);

(73) if the frame number j is L_GOPThen let j equal 1 and jump to step 2).

Although the invention has been described and illustrated with reference to specific embodiments, it is not intended that the invention be limited to these described embodiments. The present invention is described by way of example in the international video coding standard h.264, but is not limited thereto, and all equivalent changes and modifications made within the scope of the claims of the present invention are within the scope of patent protection.

Claims

1. A video code rate control method based on rho domain is characterized in that a target code rate is given

The encoding end sequentially performs a bit rate allocation algorithm of an image group level and a frame level based on a rho domain model to realize code rate control, and the method specifically comprises the following steps:

1) pre-coding a group of images, and acquiring the initial characteristic parameters of the code rate control method:

encoding video with initial quantization parameter QPA group of GOP frames of the sequence, and recording the characteristic parameter of each frame in an information base_i，ρ_i，θ_i，Δc_i，s_iIn which the letter subscript i ∈ [1, L ]_GOP]∩Z⁺Indicating the code frame number, Z, in the present GOP⁺Represents taking a positive integer; delta_i，ρ_i，θ_i，Δc_iResidual texture complexity parameters, zero coefficient proportion parameters, code rate model slope parameters and code rate model data interception difference parameters in a rho domain model corresponding to the ith frame; s_iEncoding bit rate R for motion vectors in ith frame_j ^mvAnd entropy coding bit rate R after residual coefficient quantization_j ^coeThe ratio of (A) to (B);

2) calculating a GOP bit rate budget according to the target code rate:

(21) calculating the bit rate that a GOP can be allocated to

Updating remaining available bit rate

Initialization is zero; wherein L is_GOPThe GOP length is defined, and f is a video coding frame rate;

(22) if the video frame type is I frame, skipping to step 3);

if the frame type is P frame or B frame, jumping to step 4);

3) according to the remaining available bit rate

4) according to the remaining available bit rate

5) estimating quantization coefficients QP of the current frame according to the desired zero coefficient ratio rho_j：

6) with quantization parameter QP_jEncoding the current video frame:

(61) encoding a video frame to obtain a true output bit rate for the frame

；

(62) Calculating a remaining available bit rate

Wherein the symbol max {. cndot } represents the maximum of the two numbers;

(71) storing characteristic parameters { delta, rho, theta, delta c, s } of the current frame in an information base;

(72) if frame number j ≠ L_GOPIf j is j +1, go to step 4);

(73) if the frame number j is L_GOPThen let j equal 1 and jump to step 2).

2. The method as claimed in claim 1, wherein the step 3) is performed according to the remaining available bit rate

Calculating an expected I-frame bit rate

And a corresponding zero coefficient ratio p, comprising the steps of:

(31) calculating an expected I-frame bit rate

Wherein, w_I，w_P，w_BDesired bit rate weights, 1, γ, for single frame video of type I, P and B, respectively_P，γ_BThe number of GOPs for which they occupy;

(32) updating the characteristic parameters { slope theta and intercept difference delta c } of the current I frame, and using the characteristic parameters { slope theta and intercept difference delta c } of the previous I frame adjacent to the current I frame as the characteristic parameters { slope theta and intercept difference delta c } of the current frame;

(33) calculating an expected zero coefficient ratio

Jumping to step 5); wherein,

is the bit rate of the video frame header.

3. The method as claimed in claim 1, wherein the step 4) is performed according to the remaining available bit rate

(41) calculating an available bit rate of the current frame j;

(411) if the available bit rate remains

Performing frame skipping processing;

(412) if the available bit rate remains

Then

the number of P frames in the GOP waiting for allocated bit rate minus 1:

γ_P＝γ_P-1；

the number of B frames in the GOP waiting for allocated bit rate minus 1:

γ_B＝γ_B-1；

(43) calculating the desired bit rate of the current frame

Limiting the corresponding zero coefficient ratio:

4. the method as claimed in claim 3, wherein said step (42) of detecting the characteristic parameters of the previous homogeneous frame and updating the characteristic parameters { slope θ, intercept difference Δ c and bit rate ratio s } of the current frame comprises the steps of:

(421) taking N nearest video frames of the same type before the current frame as prediction reference frames, and taking N groups of characteristic parameter values { theta ] corresponding to the N nearest video frames of the same type from an information base_i，δ_i，Δc_i，s_i1 ≦ i ≦ N, and is converted to { ln θ ≦ N_i，δ_i，Δc_i，s_i}; wherein N is a positive integer, and the subscript i belongs to Z⁺Indicating adjacent reference frame numbers; theta_iLinear rate model slope, delta, for the ith reference frame_iResidual texture complexity for the ith reference frame;

(422) linearize the slope-texture complexity model as:

lnθ(δ)＝2lnσ+α(1-δ)

(a) estimating the slope-texture complexity model parameter value of the current coding frame:

wherein,

rho-domain bit rate model slope value theta-sigma²eα^(1-δ)；

Rho-domain bit rate model truncation data difference