CN110677654A

CN110677654A - Quantization parameter cascade method of high-efficiency video coding standard low-delay coding structure

Info

Publication number: CN110677654A
Application number: CN201910980644.4A
Authority: CN
Inventors: 公衍超; 杨楷芳; 刘颖; 林庆帆; 王富平; 卢津; 王玲
Original assignee: Shaanxi Normal University; Xian University of Posts and Telecommunications
Current assignee: Shaanxi Normal University; Xian University of Posts and Telecommunications
Priority date: 2019-10-16
Filing date: 2019-10-16
Publication date: 2020-01-10
Anticipated expiration: 2039-10-16
Also published as: CN110677654B

Abstract

A quantization parameter cascade method of high-efficiency video coding standard low-delay coding structure is disclosed, which comprises determining quantization parameter of 1 st image, determining standard deviation of 1 st image, determining model parameter corresponding to 1 st image, determining 1 st image and G th image_sInter-frame difference of +1 picture, determining model parameter of 1 st temporal layer picture, determining quantization parameter offset of 1 st temporal layer picture, determining model parameter of 2 nd temporal layer picture, determining quantization parameter offset of 2 nd temporal layer picture, determining quantization parameter greater than or equal to 1 st temporal layer picture, and encoding video composition. The method solves the problems that the video time domain dependency is not considered in the existing technology aiming at the low-delay coding structure, and the acquisition of model parameters is complex. The invention has the advantages of high coding rate distortion performance, simple model parameter calculation and the likeThe method has the advantages of being applicable to video coding of a low-delay coding structure.

Description

Quantization parameter cascade method of high-efficiency video coding standard low-delay coding structure

Technical Field

The invention belongs to the technical field of video coding, and particularly relates to a quantization parameter cascading method of a high-efficiency video coding standard low-delay coding structure.

Background

Video coding is a key technology for ensuring the effective operation of a video multimedia communication system. Video data needs to be transmitted quickly in real time in video monitoring and video conference systems. However, the amount of video data collected by the camera is very large, for example, an original video in YUV420 format with a spatial resolution of 1080p and a frame rate of 50 frames/second generates 148.32MB of data in one second. Such bulky raw video cannot be quickly transmitted in real-time in current multimedia systems without encoding. The purpose of video coding is to represent the content information of the original video with fewer coding bits on the premise of satisfying the quality of the reconstructed video.

With the great popularization of high-definition devices, high-efficiency video coding standards for high-definition videos are widely adopted at present. The high-efficiency video coding standard supports three coding structures, namely a full-frame intra coding structure, a random access coding structure and a low-delay coding structure. The low-delay coding structure is suitable for scenes with high requirements on video transmission real-time performance, such as video monitoring, video conferences and the like. When the video is coded by adopting a low-delay coding structure, images in the video can be distributed to different time layers, and the coding efficiency of the video is closely related to the value of a quantization parameter selected for the images of each time layer. Quantization parameter cascading methods are a class of techniques for studying how to select the optimal quantization parameter value for each temporal layer image.

At present, the high-efficiency video coding standard adopts a fixed quantization parameter cascading method, namely, the quantization parameter of the current time layer image is added with 1 on the basis of the quantization parameter of the previous time layer image. The fixed quantization parameter cascading method does not consider the content characteristics contained in the video and the time-domain dependency between video images, so the encoding performance of the method is low. Some more efficient quantization parameter concatenation methods are proposed, which can be divided into two broad categories: a method based on experimental statistics and a method based on time-domain dependency modeling. The method based on experimental statistics is that a relation model between some video content characteristics and quantization parameter values is established by selecting a large number of videos and analyzing experimental data, for example, a model of the relation between video space-time complexity and quantization parameter values is established. Although the method based on experimental statistics considers the content characteristics of the video, the temporal dependency of the video is not considered, so the coding rate distortion performance of the method is to be improved. The method based on the time domain dependency modeling considers the video content characteristics and the time domain dependency among video images at the same time, and the method can achieve higher coding rate distortion performance theoretically. However, most of the methods proposed currently are proposed for random access coding structures. Since there is a significant difference in the prediction relationship between the random access coding structure and the low-latency coding structure, these methods are not suitable for the low-latency coding structure. In addition, most methods have complex model parameter obtaining modes, such as pre-coding and motion estimation, and the complex calculation process is not suitable for real-time video application of a low-delay coding structure.

Disclosure of Invention

The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art, and provide a quantization parameter cascading method for a high-efficiency video coding standard low-delay coding structure, which effectively considers the video content characteristics and the time-domain dependency, and has simple model parameter calculation.

The technical scheme adopted for solving the technical problems comprises the following steps:

(1) determining quantization parameter of 1 st image

Quantization parameter QP for a first picture_ISet by the operator in the configuration file, QP_I∈{1,2,...,51}。

(2) Determining the standard deviation of the 1 st image

The spatial resolution of the input video image is obtained by obtaining the side length b of the basic calculation unit according to the formula (1)_s：

Where round () is the rounding function, w is the image width, h is the image height, p₁∈[0.5,3.5]，p₂∈[0.05,0.4]，w_b∈{88,89,...,704}，h_b∈{72,73,...,576}。

The basic calculation unit being a side length of b_sSquare pixel block of (1), standard deviation sd [1 ] of the 1 st image]Determined according to equation (2):

where int () is the lower integer function, y [ n, u,1]The luminance value of the nth pixel of the u-th basic computing unit in the 1 st image is obtained, u and n are finite positive integers, b_s ²Is the number of pixels of the square pixel block.

(3) Determining the model parameter corresponding to the 1 st image

Model parameter alpha corresponding to the 1 st image₀，β₀，γ₀，σ₀Determined according to equations (3) - (6):

α₀＝ξ₁sd[1]+ε₁(3)

β₀＝ξ₂sd[1]+ε₂(4)

γ₀＝ξ₃sd[1]+ε₃(5)

σ₀＝ξ₄sd[1]+ε₄(6)

in which ξ₁∈[0.01,0.06]，ξ₂∈[0.01,0.05]，ξ₃∈[0.1,0.5]，ξ₄∈[0.0005,0.01]，ε₁∈[0.1,1]，ε₂∈[0.1,0.7]，ε₃∈[-1.8,-1]，ε₄∈[-1,-0.1]。

(4) Determining the 1 st image and the G th image_sInter-frame difference of +1 images

The basic calculation unit being a side length of b_sSquare pixel block, 1 st image and G_sFrame-to-frame difference ld [ G ] of +1 images_s+1]Determined according to equation (7):

wherein G is_sIs the size of the group of images.

(5) Determining model parameters for 1 st temporal layer images

Model parameter mu corresponding to 1 st time layer image_(a,1)，μ_(b,1)，μ_(c,1)，η_(c,1)，μ_(d,1)，η_(d,1)Determined according to equations (8) - (13):

μ_(a,1)＝ω_1,1ld[G_s+1]+ψ_1,1(8)

μ_(b,1)＝ω_1,2ld[G_s+1]+ψ_1,2(9)

μ_(c,1)＝ω_1,3ld[G_s+1]+ψ_1,3(10)

η_(c,1)＝ω_1,4ld[G_s+1]+ψ_1,4(11)

μ_(d,1)＝ω_1,5ld[G_s+1]+ψ_1,5(12)

η_(d,1)＝ω_1,6ld[G_s+1]+ψ_1,6(13)

wherein ω is_1,1∈[0.001,0.012]，ω_1,2∈[-0.2,0]，ω_1,3∈[0.0005,0.003]，ω_1,4∈[0.001,0.02]，ω_1,5∈[-0.002,0]，ω_1,6∈[0,0.012]，ψ_1,1∈[0.01,0.2]，ψ_1,2∈[0.1,2]，ψ_1,3∈[-0.2,0]，ψ_1,4∈[0,2]，ψ_1,5∈[0,0.01]，ψ_1,6∈[-0.3,0]。

(6) Determining quantization parameter offset for 1 st temporal layer picture

Quantization parameter offset x corresponding to 1 st temporal layer image_1,optDetermined according to equation (14):

wherein a is [1,10 ]]，b∈[5,21]，θ₁∈[1,8]。

(7) Determining model parameters for 2 nd temporal layer images

Model parameter mu corresponding to 2 nd time layer image_(a,2)，μ_(b,2)，μ_(c,2)，η_(c,2)，μ_(d,2)，η_(d,2)Determined according to equations (15) - (20):

μ_(a,2)＝ω_2,1ld[G_s+1]+ψ_2,1(15)

μ_(b,2)＝ω_2,2ld[G_s+1]+ψ_2,2(16)

μ_(c,2)＝ω_2,3ld[G_s+1]+ψ_2,3(17)

η_(c,2)＝ω_2,4ld[G_s+1]+ψ_2,4(18)

μ_(d,2)＝ω_2,5ld[G_s+1]+ψ_2,5(19)

η_(d,2)＝ω_2,6ld[G_s+1]+ψ_2,6(20)

wherein ω is_2,1∈[0.001,0.012]，ω_2,2∈[-0.2,0]，ω_2,3∈[0.0005,0.003]，ω_2,4∈[0.0001,0.002]，ω_2,5∈[-0.002,0]，ω_2,6∈[0,0.012]，ψ_2,1∈[0.01,0.2]，ψ_2,2∈[0.1,2]，ψ_2,3∈[-0.2,0]，ψ_2,4∈[0,2]，ψ_2,5∈[0,0.01]，ψ_2,6∈[-0.3,0]。

(8) Determining quantization parameter offset for 2 nd temporal layer picture

2 nd temporal layer imageOffset x of quantization parameter_2,optDetermined according to equation (21):

wherein theta is₂∈[1,15]。

(9) Determining quantization parameters greater than or equal to 1 st temporal layer image

The quantization parameter at greater than or equal to 1 st temporal layer picture is determined as follows (22):

wherein clip3 (k)₁|k₂≤k₁≤k₃) For restricted functions, return k₁Value, and k₁∈[k₂,k₃]，QP_LRepresenting the quantization parameter, Δ, of the L-th temporal layer₁∈[2,15]，Δ₂∈[2,8]。

(10) Encoding video

Using the coding parameter configuration file corresponding to the low-delay coding structure and determining the QP according to the step (1)_ICoding the 1 st picture of the video according to the QP determined in step (9)_LL is more than or equal to 1, and other images in the 1 st time layer in the coded video are larger than or equal to the 1 st time layer.

In the step (2) of determining the standard deviation of the 1 st image, p is₁Most preferably 2.3120, p₂Most preferably 0.1098, w_bMost preferably 176, h_bAnd most preferably 144.

In the step (3) of determining the model parameters corresponding to the 1 st image, the xi is₁Optimally 0.0323 ξ₂Optimally 0.0224, ξ₃Optimally 0.3254 ξ₄Most preferably 0.006,. epsilon₁Most preferably 0.5026,. epsilon₂Most preferably 0.3258,. epsilon₃Most preferably-1.6644,. epsilon₄Most preferably-0.5214.

In the step (5) of determining the model parameters of the 1 st temporal layer image, ω is_1,1Most preferably 0.0088, omega_1,2Most preferably-0.1058, omega_1,3Most preferably 0.0016, omega_1,4Most preferably 0.0068, omega_1,5Most preferably-0.0009, omega_1,6Most preferably 0.002, psi_1,1Most preferably 0.0751,. psi_1,2Most preferably 0.933, psi_1,3Most preferably-0.0610, psi_1,4Most preferably 0.3127, psi_1,5Most preferably 0.0052,. psi_1,6Most preferably-0.0365.

In the step (6) of determining the quantization parameter offset of the 1 st temporal layer image, a is preferably 4.2005, b is preferably 13.7122, and θ is preferably set as₁And most preferably 2.

In the step (7) of determining the model parameters of the 2 nd temporal layer image, the parameter ω is_2,1Most preferably 0.0065, omega_2,2Most preferably-0.0348, omega_2,3Most preferably 0.0006, omega_2,4Most preferably 0.0005, omega_2,5Most preferably-0.0001, omega_2,6Preferably 0.0012,. psi_2,1Most preferably 0.0202, psi_2,2Most preferably 0.3282, psi_2,3Most preferably-0.0405, psi_2,4Most preferably 0.2079, psi_2,5Preferably 0.0019,. psi_2,6The most preferable range is-0.0143.

In the step (8) of determining quantization parameter offset of 2 nd time layer image, the value theta is₂And most preferably 9.

In the step (9) of determining the quantization parameter greater than or equal to the 1 st temporal layer image, the value Δ is₁Most preferably 8, Δ₂Most preferably 3.

The invention adopts the steps of determining the standard deviation of the 1 st image and determining the 1 st image and the G th image_sThe step of inter-frame difference of +1 image considers the influence of the video content characteristics on the quantization parameter, and determines the 1 st image and the G th image_sThe inter-frame difference step of +1 image, the model parameter step of determining the 1 st time layer image and the model parameter step of determining the 2 nd time layer image are adopted to consider the influence of time-domain dependency on quantization parameters, and the model parameter step of determining the 1 st image, the model parameter step of determining the 1 st time layer image and the model parameter step of determining the 2 nd time layer image are adopted, so that the method can be used for simply and quickly determining the inter-frame difference of the 1 st imageThe method solves the problems that the video time domain dependency is not considered in the existing technology aiming at the low-delay coding structure and the model parameter is complex to obtain. The invention has the advantages of high coding rate distortion performance, simple model parameter calculation and the like, and can be used for video coding of a low-delay coding structure.

Drawings

Fig. 1 is a flowchart of a quantization parameter concatenation method of a low-latency coding structure of an efficient video coding standard according to embodiment 1.

Detailed Description

The present invention will be described in further detail with reference to the drawings and examples, but the present invention is not limited to the examples.

Example 1

The quantization parameter cascading method of the high-efficiency video coding standard low-delay coding structure in the embodiment comprises the following steps:

(1) determining quantization parameter of 1 st image

(2) Determining the standard deviation of the 1 st image

Where round () is the rounding function, w is the image width, h is the image height, p₁∈[0.5,3.5]，p₂∈[0.05,0.4]，w_b∈{88,89,...,704}，h_bE {72, 73.., 576 }. P of the present example₁Is 2.3120, p₂Is 0.1098, w_bIs 176, h_bIs 144.

(3) Determining the model parameter corresponding to the 1 st image

α₀＝ξ₁sd[1]+ε₁(3)

β₀＝ξ₂sd[1]+ε₂(4)

γ₀＝ξ₃sd[1]+ε₃(5)

σ₀＝ξ₄sd[1]+ε₄(6)

in which ξ₁∈[0.01,0.06]，ξ₂∈[0.01,0.05]，ξ₃∈[0.1,0.5]，ξ₄∈[0.0005,0.01]，ε₁∈[0.1,1]，ε₂∈[0.1,0.7]，ε₃∈[-1.8,-1]，ε₄∈[-1,-0.1]. Xi of the present embodiment₁0.0323, ξ₂Is 0.0224, ξ₃0.3254, ξ₄Is 0.006,. epsilon₁Is 0.5026, ε₂Is 0.3258, ε₃Is-1.6644, ε₄Is-0.5214.

wherein G is_sIs the size of the group of images.

(5) Determining model parameters for 1 st temporal layer images

1 st oneModel parameter mu corresponding to time-layer image_(a,1)，μ_(b,1)，μ_(c,1)，η_(c,1)，μ_(d,1)，η_(d,1)Determined according to equations (8) - (13):

μ_(a,1)＝ω_1,1ld[G_s+1]+ψ_1,1(8)

μ_(b,1)＝ω_1,2ld[G_s+1]+ψ_1,2(9)

μ_(c,1)＝ω_1,3ld[G_s+1]+ψ_1,3(10)

η_(c,1)＝ω_1,4ld[G_s+1]+ψ_1,4(11)

μ_(d,1)＝ω_1,5ld[G_s+1]+ψ_1,5(12)

η_(d,1)＝ω_1,6ld[G_s+1]+ψ_1,6(13)

wherein ω is_1,1∈[0.001,0.012]，ω_1,2∈[-0.2,0]，ω_1,3∈[0.0005,0.003]，ω_1,4∈[0.001,0.02]，ω_1,5∈[-0.002,0]，ω_1,6∈[0,0.012]，ψ_1,1∈[0.01,0.2]，ψ_1,2∈[0.1,2]，ψ_1,3∈[-0.2,0]，ψ_1,4∈[0,2]，ψ_1,5∈[0,0.01]，ψ_1,6∈[-0.3,0]. ω of the present embodiment_1,1Is 0.0088, omega_1,2Is-0.1058, omega_1,3Is 0.0016, omega_1,4Is 0.0068, omega_1,5Is-0.0009, omega_1,6Is 0.002, psi_1,1Is 0.0751, psi_1,2Is 0.933,. psi_1,3Is-0.0610, psi_1,4Is 0.3127, psi_1,5Is 0.0052,. psi_1,6Is-0.0365.

(6) Determining quantization parameter offset for 1 st temporal layer picture

wherein a is [1,10 ]]，b∈[5,21]，θ₁∈[1,8]. In this example, a is 4.2005, b is 13.7122, and θ₁Is 2.

(7) Determining model parameters for 2 nd temporal layer images

μ_(a,2)＝ω_2,1ld[G_s+1]+ψ_2,1(15)

μ_(b,2)＝ω_2,2ld[G_s+1]+ψ_2,2(16)

μ_(c,2)＝ω_2,3ld[G_s+1]+ψ_2,3(17)

η_(c,2)＝ω_2,4ld[G_s+1]+ψ_2,4(18)

μ_(d,2)＝ω_2,5ld[G_s+1]+ψ_2,5(19)

η_(d,2)＝ω_2,6ld[G_s+1]+ψ_2,6(20)

wherein ω is_2,1∈[0.001,0.012]，ω_2,2∈[-0.2,0]，ω_2,3∈[0.0005,0.003]，ω_2,4∈[0.0001,0.002]，ω_2,5∈[-0.002,0]，ω_2,6∈[0,0.012]，ψ_2,1∈[0.01,0.2]，ψ_2,2∈[0.1,2]，ψ_2,3∈[-0.2,0]，ψ_2,4∈[0,2]，ψ_2,5∈[0,0.01]，ψ_2,6∈[-0.3,0]. ω of the present embodiment_2,1Is 0.0065, omega_2,2Is-0.0348, omega_2,3Is 0.0006, omega_2,4Is 0.0005, omega_2,5Is-0.0001, omega_2,6Is 0.0012,. psi_2,1Is 0.0202, psi_2,2Is 0.3282, psi_2,3Is-0.0405, psi_2,4Is 0.2079, psi_2,5Is 0.0019,. psi_2,6It was-0.0143.

(8) Determining quantization parameter offset for 2 nd temporal layer picture

Quantization parameter corresponding to 2 nd time layer imageNumber offset x_2,optDetermined according to equation (21):

wherein theta is₂∈[1,15]. Theta of the embodiment₂Is 9.

wherein clip3 (k)₁|k₂≤k₁≤k₃) For restricted functions, return k₁Value, and k₁∈[k₂,k₃]，QP_LRepresenting the quantization parameter, Δ, of the L-th temporal layer₁∈[2,15]，Δ₂∈[2,8]. Delta of the embodiment₁Is 8, Δ₂Is 3.

(10) Encoding video

And finishing the quantization parameter cascading method of the high-efficiency video coding standard low-delay coding structure.

Example 2

(1) determining quantization parameter of 1 st image

This procedure is the same as in example 1.

(2) Determining the standard deviation of the 1 st image

Where round () is the rounding function, w is the image width, h is the image height, p₁∈[0.5,3.5]，p₂∈[0.05,0.4]，w_b∈{88,89,...,704}，h_bE {72, 73.., 576 }. P of the present example₁Is 0.5, p₂Is 0.05, w_bIs 88, h_bIs 72.

(3) Determining the model parameter corresponding to the 1 st image

α₀＝ξ₁sd[1]+ε₁(3)

β₀＝ξ₂sd[1]+ε₂(4)

γ₀＝ξ₃sd[1]+ε₃(5)

σ₀＝ξ₄sd[1]+ε₄(6)

in which ξ₁∈[0.01,0.06]，ξ₂∈[0.01,0.05]，ξ₃∈[0.1,0.5]，ξ₄∈[0.0005,0.01]，ε₁∈[0.1,1]，ε₂∈[0.1,0.7]，ε₃∈[-1.8,-1]，ε₄∈[-1,-0.1]. Xi of the present embodiment₁Is 0.01, ξ₂Is 0.01, ξ₃Is 0.1, ξ₄Is 0.0005,. epsilon₁The content of the organic acid is 0.1,ε₂is 0.1,. epsilon₃Is-1.8,. epsilon₄Is-1.

This procedure is the same as in example 1.

(5) Determining model parameters for 1 st temporal layer images

μ_(a,1)＝ω_1,1ld[G_s+1]+ψ_1,1(8)

μ_(b,1)＝ω_1,2ld[G_s+1]+ψ_1,2(9)

μ_(c,1)＝ω_1,3ld[G_s+1]+ψ_1,3(10)

η_(c,1)＝ω_1,4ld[G_s+1]+ψ_1,4(11)

μ_(d,1)＝ω_1,5ld[G_s+1]+ψ_1,5(12)

η_(d,1)＝ω_1,6ld[G_s+1]+ψ_1,6(13)

wherein ω is_1,1∈[0.001,0.012]，ω_1,2∈[-0.2,0]，ω_1,3∈[0.0005,0.003]，ω_1,4∈[0.001,0.02]，ω_1,5∈[-0.002,0]，ω_1,6∈[0,0.012]，ψ_1,1∈[0.01,0.2]，ψ_1,2∈[0.1,2]，ψ_1,3∈[-0.2,0]，ψ_1,4∈[0,2]，ψ_1,5∈[0,0.01]，ψ_1,6∈[-0.3,0]. ω of the present embodiment_1,1Is 0.001, omega_1,2Is-0.2, omega_1,3Is 0.0005, omega_1,4Is 0.001, omega_1,5Is-0.002, omega_1,6Is 0, psi_1,1Is 0.01, #_1,2Is 0.1, psi_1,3Is-0.2, #_1,4Is 0, psi_1,5Is 0, psi_1,6Is-0.3.

(6) Determining quantization parameter offset for 1 st temporal layer picture

wherein a is [1,10 ]]，b∈[5,21]，θ₁∈[1,8]. In this example, a is 1, b is 5, and θ₁Is 1.

(7) Determining model parameters for 2 nd temporal layer images

μ_(a,2)＝ω_2,1ld[G_s+1]+ψ_2,1(15)

μ_(b,2)＝ω_2,2ld[G_s+1]+ψ_2,2(16)

μ_(c,2)＝ω_2,3ld[G_s+1]+ψ_2,3(17)

η_(c,2)＝ω_2,4ld[G_s+1]+ψ_2,4(18)

μ_(d,2)＝ω_2,5ld[G_s+1]+ψ_2,5(19)

η_(d,2)＝ω_2,6ld[G_s+1]+ψ_2,6(20)

wherein ω is_2,1∈[0.001,0.012]，ω_2,2∈[-0.2,0]，ω_2,3∈[0.0005,0.003]，ω_2,4∈[0.0001,0.002]，ω_2,5∈[-0.002,0]，ω_2,6∈[0,0.012]，ψ_2,1∈[0.01,0.2]，ψ_2,2∈[0.1,2]，ψ_2,3∈[-0.2,0]，ψ_2,4∈[0,2]，ψ_2,5∈[0,0.01]，ψ_2,6∈[-0.3,0]. ω of the present embodiment_2,1Is 0.001, omega_2,2Is-0.2, omega_2,3Is 0.0005, omega_2,4Is 0.0001, omega_2,5Is-0.002, omega_2,6Is 0, psi_2,1Is 0.01, #_2,2Is 0.1, psi_2,3Is-0.2, #_2,4Is 0, psi_2,5Is 0, psi_2,6Is-0.3.

(8) Determining quantization parameter offset for 2 nd temporal layer picture

Quantization parameter offset x corresponding to 2 nd temporal layer image_2,optDetermined according to equation (21):

wherein theta is₂∈[1,15]. Theta of the embodiment₂Is 1.

wherein clip3 (k)₁|k₂≤k₁≤k₃) For restricted functions, return k₁Value, and k₁∈[k₂,k₃]，QP_LRepresenting the quantization parameter, Δ, of the L-th temporal layer₁∈[2,15]，Δ₂∈[2,8]. Delta of the embodiment₁Is 2, Δ₂Is 2.

(10) Encoding video

This procedure is the same as in example 1.

Example 3

(1) determining quantization parameter of 1 st image

This procedure is the same as in example 1.

(2) Determining the standard deviation of the 1 st image

Spatial segmentation of input video imagesResolution, obtaining the side length b of the basic calculation unit according to equation (1)_s：

Where round () is the rounding function, w is the image width, h is the image height, p₁∈[0.5,3.5]，p₂∈[0.05,0.4]，w_b∈{88,89,...,704}，h_bE {72, 73.., 576 }. P of the present example₁Is 3.5, p₂Is 0.4, w_bIs 704, h_bIs 576.

(3) Determining the model parameter corresponding to the 1 st image

α₀＝ξ₁sd[1]+ε₁(3)

β₀＝ξ₂sd[1]+ε₂(4)

γ₀＝ξ₃sd[1]+ε₃(5)

σ₀＝ξ₄sd[1]+ε₄(6)

in which ξ₁∈[0.01,0.06]，ξ₂∈[0.01,0.05]，ξ₃∈[0.1,0.5]，ξ₄∈[0.0005,0.01]，ε₁∈[0.1,1]，ε₂∈[0.1,0.7]，ε₃∈[-1.8,-1]，ε₄∈[-1,-0.1]. Xi of the present embodiment₁Is 0.06, ξ₂Is 0.05, ξ₃Is 0.5, ξ₄Is 0.01 epsilon₁Is 1, epsilon₂Is 0.7,. epsilon₃Is-1, epsilon₄Is-0.1.

This procedure is the same as in example 1.

(5) Determining model parameters for 1 st temporal layer images

μ_(a,1)＝ω_1,1ld[G_s+1]+ψ_1,1(8)

μ_(b,1)＝ω_1,2ld[G_s+1]+ψ_1,2(9)

μ_(c,1)＝ω_1,3ld[G_s+1]+ψ_1,3(10)

η_(c,1)＝ω_1,4ld[G_s+1]+ψ_1,4(11)

μ_(d,1)＝ω_1,5ld[G_s+1]+ψ_1,5(12)

η_(d,1)＝ω_1,6ld[G_s+1]+ψ_1,6(13)

wherein ω is_1,1∈[0.001,0.012]，ω_1,2∈[-0.2,0]，ω_1,3∈[0.0005,0.003]，ω_1,4∈[0.001,0.02]，ω_1,5∈[-0.002,0]，ω_1,6∈[0,0.012]，ψ_1,1∈[0.01,0.2]，ψ_1,2∈[0.1,2]，ψ_1,3∈[-0.2,0]，ψ_1,4∈[0,2]，ψ_1,5∈[0,0.01]，ψ_1,6∈[-0.3,0]. ω of the present embodiment_1,1Is 0.012, omega_1,2Is 0, omega_1,3Is 0.003, omega_1,4Is 0.02, omega_1,5Is 0, omega_1,6Is 0.012, psi_1,1Is 0.2, #_1,2Is 2, psi_1,3Is 0, psi_1,4Is 2, psi_1,5Is 0.01, #_1,6Is 0.

(6) Determining quantization parameter offset for 1 st temporal layer picture

wherein a is [1,10 ]]，b∈[5,21]，θ₁∈[1,8]. In this example, a is 10, b is 21, and θ₁Is 8.

(7) Determining model parameters for 2 nd temporal layer images

μ_(a,2)＝ω_2,1ld[G_s+1]+ψ_2,1(15)

μ_(b,2)＝ω_2,2ld[G_s+1]+ψ_2,2(16)

μ_(c,2)＝ω_2,3ld[G_s+1]+ψ_2,3(17)

η_(c,2)＝ω_2,4ld[G_s+1]+ψ_2,4(18)

μ_(d,2)＝ω_2,5ld[G_s+1]+ψ_2,5(19)

η_(d,2)＝ω_2,6ld[G_s+1]+ψ_2,6(20)

wherein ω is_2,1∈[0.001,0.012]，ω_2,2∈[-0.2,0]，ω_2,3∈[0.0005,0.003]，ω_2,4∈[0.0001,0.002]，ω_2,5∈[-0.002,0]，ω_2,6∈[0,0.012]，ψ_2,1∈[0.01,0.2]，ψ_2,2∈[0.1,2]，ψ_2,3∈[-0.2,0]，ψ_2,4∈[0,2]，ψ_2,5∈[0,0.01]，ψ_2,6∈[-0.3,0]. ω of the present embodiment_2,1Is 0.012, omega_2,2Is 0, omega_2,3Is 0.003, omega_2,4Is 0.002, omega_2,5Is a non-volatile organic compound (I) with a value of 0,ω_2,6is 0.012, psi_2,1Is 0.2, #_2,2Is 2, psi_2,3Is 0, psi_2,4Is 2, psi_2,5Is 0.01, #_2,6Is 0.

(8) Determining quantization parameter offset for 2 nd temporal layer picture

wherein theta is₂∈[1,15]. Theta of the embodiment₂Is 15.

wherein clip3 (k)₁|k₂≤k₁≤k₃) For restricted functions, return k₁Value, and k₁∈[k₂,k₃]，QP_LRepresenting the quantization parameter, Δ, of the L-th temporal layer₁∈[2,15]，Δ₂∈[2,8]. Delta of the embodiment₁Is 15, Δ₂Is 8.

(10) Encoding video

This procedure is the same as in example 1.

To verify the beneficial effects of the present invention, the inventor performed experiments on test videos by using the method of embodiment 1 of the present invention, the experiments are as follows:

the method comprises the steps of selecting 20 videos of 5 types recommended by a high-definition video coding standard as test videos, selecting an encoder HM16.0 recommended by an organization and made by the high-definition video coding standard, adopting a low-delay coding structure, wherein all frames are P frames, and the size of an image group is 4.

The method of embodiment 1 of the invention and the fixed quantization parameter cascading method are adopted to respectively encode 20 videos, and BD-rate indexes are used for measuring the rate distortion performance of the encoded videos corresponding to the two methods, wherein the BD-rate is disclosed in An exceld-in for computing Bjontegaard measurement and bits evaluation, VCEG-AE07, Marrakech, MA, Jan.2007. And for each test video, taking the coding distortion and code rate corresponding to the fixed quantization parameter cascading method as reference values to obtain the BD-rate value corresponding to the method provided by the invention. When the BD-rate is negative, the code rate is reduced and the coding rate distortion performance is improved under the same quality.

The results obtained according to the process of example 1 of the present invention are shown in Table 1.

TABLE 1 BD-rate (%) values corresponding to the method of the invention

As can be seen from table 1, the average BD-rate for all tested videos is-3.99% for the proposed method compared to the fixed quantization parameter cascade method. Compared with the fixed quantization parameter cascading method, the method provided by the invention can save the code rate by 3.99% on the premise of achieving the same quality of the coded and reconstructed video, and the coding rate distortion performance is higher.

Claims

1. A quantization parameter cascading method of a high-efficiency video coding standard low-delay coding structure is characterized by comprising the following steps:

(1) determining quantization parameter of 1 st image

Quantization parameter QP for a first picture_ISet by the operator in the configuration file, QP_I∈{1,2,...,51}；

(2) Determining the standard deviation of the 1 st image

Where round () is the rounding function, w is the image width, h is the image height, p₁∈[0.5,3.5]，p₂∈[0.05,0.4]，w_b∈{88,89,...,704}，h_b∈{72,73,...,576}；

where int () is the lower integer function, y [ n, u,1]The luminance value of the nth pixel of the u-th basic computing unit in the 1 st image is obtained, u and n are finite positive integers, b_s ²The number of pixels of a square pixel block;

(3) determining the model parameter corresponding to the 1 st image

α₀＝ξ₁sd[1]+ε₁(3)

β₀＝ξ₂sd[1]+ε₂(4)

γ₀＝ξ₃sd[1]+ε₃(5)

σ₀＝ξ₄sd[1]+ε₄(6)

in which ξ₁∈[0.01,0.06]，ξ₂∈[0.01,0.05]，ξ₃∈[0.1,0.5]，ξ₄∈[0.0005,0.01]，ε₁∈[0.1,1]，ε₂∈[0.1,0.7]，ε₃∈[-1.8,-1]，ε₄∈[-1,-0.1]；

wherein G is_sIs the size of the group of images.

(5) Determining model parameters for 1 st temporal layer images

μ_(a,1)＝ω_1,1ld[G_s+1]+ψ_1,1(8)

μ_(b,1)＝ω_1,2ld[G_s+1]+ψ_1,2(9)

μ_(c,1)＝ω_1,3ld[G_s+1]+ψ_1,3(10)

η_(c,1)＝ω_1,4ld[G_s+1]+ψ_1,4(11)

μ_(d,1)＝ω_1,5ld[G_s+1]+ψ_1,5(12)

η_(d,1)＝ω_1,6ld[G_s+1]+ψ_1,6(13)

wherein ω is_1,1∈[0.001,0.012]，ω_1,2∈[-0.2,0]，ω_1,3∈[0.0005,0.003]，ω_1,4∈[0.001,0.02]，ω_1,5∈[-0.002,0]，ω_1,6∈[0,0.012]，ψ_1,1∈[0.01,0.2]，ψ_1,2∈[0.1,2]，ψ_1,3∈[-0.2,0]，ψ_1,4∈[0,2]，ψ_1,5∈[0,0.01]，ψ_1,6∈[-0.3,0]；

(6) Determining quantization parameter offset for 1 st temporal layer picture

whereina∈[1,10]，b∈[5,21]，θ₁∈[1,8]；

(7) Determining model parameters for 2 nd temporal layer images

μ_(a,2)＝ω_2,1ld[G_s+1]+ψ_2,1(15)

μ_(b,2)＝ω_2,2ld[G_s+1]+ψ_2,2(16)

μ_(c,2)＝ω_2,3ld[G_s+1]+ψ_2,3(17)

η_(c,2)＝ω_2,4ld[G_s+1]+ψ_2,4(18)

μ_(d,2)＝ω_2,5ld[G_s+1]+ψ_2,5(19)

η_(d,2)＝ω_2,6ld[G_s+1]+ψ_2,6(20)

wherein ω is_2,1∈[0.001,0.012]，ω_2,2∈[-0.2,0]，ω_2,3∈[0.0005,0.003]，ω_2,4∈[0.0001,0.002]，ω_2,5∈[-0.002,0]，ω_2,6∈[0,0.012]，ψ_2,1∈[0.01,0.2]，ψ_2,2∈[0.1,2]，ψ_2,3∈[-0.2,0]，ψ_2,4∈[0,2]，ψ_2,5∈[0,0.01]，ψ_2,6∈[-0.3,0]；

(8) Determining quantization parameter offset for 2 nd temporal layer picture

wherein theta is₂∈[1,15]；

wherein clip3 (k)₁|k₂≤k₁≤k₃) For restricted functions, return k₁Value, and k₁∈[k₂,k₃]，QP_LRepresenting the quantization parameter, Δ, of the L-th temporal layer₁∈[2,15]，Δ₂∈[2,8]；

(10) Encoding video

2. The method of claim 1, wherein the quantization parameter concatenation method for a high efficiency video coding standard low latency coding structure is as follows: in the step (2) of determining the standard deviation of the 1 st image, p is₁Is 2.3120, p₂Is 0.1098, w_bIs 176, h_bIs 144.

3. The method of claim 1, wherein the quantization parameter concatenation method for a high efficiency video coding standard low latency coding structure is as follows: in the step (3) of determining the model parameter corresponding to the 1 st image, the xi is₁0.0323, ξ₂Is 0.0224, ξ₃0.3254, ξ₄Is 0.006,. epsilon₁Is 0.5026, ε₂Is 0.3258, ε₃Is-1.6644, ε₄Is-0.5214.

4. The method of claim 1, wherein the quantization parameter concatenation method for a high efficiency video coding standard low latency coding structure is as follows: in the step (5) of determining model parameters of the 1 st temporal layer image, ω is_1,1Is 0.0088, omega_1,2Is-0.1058, omega_1,3Is 0.0016, omega_1,4Is 0.0068, omega_1,5Is-0.0009, omega_1,6Is 0.002, psi_1,1Is 0.0751, psi_1,2Is 0.933,. psi_1,3Is-0.0610, psi_1,4Is 0.3127, psi_1,5Is 0.0052,. psi_1,6Is-0.0365.

5. The method of claim 1, wherein the quantization parameter concatenation method for a high efficiency video coding standard low latency coding structure is as follows: in the step (6) of determining the quantization parameter offset of the 1 st temporal layer picture, a is 4.2005, b is 13.7122, and theta is₁Is 2.

6. The method of claim 1, wherein the quantization parameter concatenation method for a high efficiency video coding standard low latency coding structure is as follows: in the step (7) of determining model parameters of the 2 nd temporal layer image, ω is_2,1Is 0.0065, omega_2,2Is-0.0348, omega_2,3Is 0.0006, omega_2,4Is 0.0005, omega_2,5Is-0.0001, omega_2,6Is 0.0012,. psi_2,1Is 0.0202, psi_2,2Is 0.3282, psi_2,3Is-0.0405, psi_2,4Is 0.2079, psi_2,5Is 0.0019,. psi_2,6It was-0.0143.

7. The method of claim 1, wherein the quantization parameter concatenation method for a high efficiency video coding standard low latency coding structure is as follows: in the step (8) of determining the quantization parameter offset of the 2 nd time layer image, the theta₂Is 9.

8. The method of claim 1, wherein the quantization parameter concatenation method for a high efficiency video coding standard low latency coding structure is as follows: in the step (9) of determining the quantization parameter greater than or equal to the 1 st temporal layer image, Δ₁Is 8, Δ₂Is 3.