Browse Prior Art Database

Bits-QP model in Rate control for video encoder

IP.com Disclosure Number: IPCOM000131140D
Original Publication Date: 2005-Dec-10
Included in the Prior Art Database: 2005-Dec-10
Document File: 2 page(s) / 33K

Publishing Venue

Siemens

Related People

Juergen Carstens: CONTACT

Abstract

International video coding standards, such as MPEG-1, 2, 4 (MPEG, Moving Picture Experts Group) and H.26x remove temporal and spatial redundancies by adopting motion-compensation, transformation and quantization. They also remove statistical redundancy by using variable length coding (VLC). Therefore, these coding schemes will generate variable bit rates. In applications where a compressed video bit stream with variable bit rates is transmitted through a limited channel bandwidth, a proper bit-rate control algorithm is necessary to smooth the bit rate variations and prevent the buffers of both encoder and decoder from overflowing and underflowing. Rate control is a decision-making process where the desired encoding rate for a source video can be met accurately by properly setting a sequence of quantization parameters (QP). Figure 1 shows a flowchart of rate control. Normally, the first target bits are allocated for the frame and the macroblock (MB). Then the quantization parameter (QP) is calculated according to model the MB and to encode it. Finally the model is updated with the current data (such as QP and bits for MB). Up to now, there are four main rate control techniques: MPEG-2 (Test model 5), H263 (Video Codec Test Model Version 8), MPEG-4 (Q2) and H264/MPEG-4 Advanced Video Coding (JM). Their QP models are different from each other and in general they all introduce complexity parameters, some based on Laplacian distribution. These complexity parameters only concern luminance signals but not chrominance signals which are also part of the MB. Moreover, there is much correlation between the two signal types in the MB and under some conditions it would get imprecise.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 53% of the total text.

Page 1 of 2

S

Bits-QP model in Rate control for video encoder

Idea: Xiang Li, DE-Munich

International video coding standards, such as MPEG-1, 2, 4 (MPEG, Moving Picture Experts Group) and H.26x remove temporal and spatial redundancies by adopting motion-compensation, transformation and quantization. They also remove statistical redundancy by using variable length coding (VLC). Therefore, these coding schemes will generate variable bit rates. In applications where a compressed video bit stream with variable bit rates is transmitted through a limited channel bandwidth, a proper bit-rate control algorithm is necessary to smooth the bit rate variations and prevent the buffers of both encoder and decoder from overflowing and underflowing. Rate control is a decision-making process where the desired encoding rate for a source video can be met accurately by properly setting a sequence of quantization parameters (QP).

Figure 1 shows a flowchart of rate control. Normally, the first target bits are allocated for the frame and the macroblock (MB). Then the quantization parameter (QP) is calculated according to model the MB and to encode it. Finally the model is updated with the current data (such as QP and bits for MB). Up to now, there are four main rate control techniques: MPEG-2 (Test model 5), H263 (Video Codec Test Model Version 8), MPEG-4 (Q2) and H264/MPEG-4 Advanced Video Coding (JM). Their QP models are different from each other and in general they all introduce complexity parameters, some based on Laplacian distribution. These complexity parameters only concern luminance signals but not chrominance signals which are also part of the MB. Moreover, there is much correlation between the two signal types in the MB and under some conditions it would get imprecise.

Considering the defects of presently known QP models, a new solution is proposed. According to statistics, the curve of average bits per macroblock against QP is approximated by a third-order Bezier curve. Figure 2 gives an example of statistics for standard test sequences.

Based on such observation, the following equations are used to describe bits/MB and QP:

⎩ ⎨ ⎧

P

BitsMB

      + - = t

0

where BitsMB is the bit-rate parameter (target bits of the MB), P0 - P3 are model parameters. According to the properties of the Bezier curve, P0 and P3 are the beginning and ending points of the curve, respectively. These two points are corresponding to the BitsMB o...