Lexicographic Bit Allocation for MPEG Video

Abstract

MPEG standards and MPEG-compliant bitstreams must be decodable by the VBV. We consider the problem of allocating bits among pictures in an MPEG video coder to equalize the visual quality of theIn this paper, we develop a novel framework for bit coded pictures, while meeting buffer and channel constraints allocation under VBV constraints and a total bit budget. imposed by the MPEG video buffering verifier. We address this This framework consists of three components: (1) a bitproblem within a framework that consists of three components: production model, (2) a novel lexicographic optimality cri-(1) a bit production model for the input pictures, (2) a set of terion, and (3) a set of buffer constraints for constant and bit-rate constraints imposed by the video buffering verifier, and variable bit-rate operation. We formalize bit allocation as (3) a novel lexicographic criterion for optimality. Under this a resource allocation problem with continuous variables framework, we derive simple necessary and sufficient condiand nonlinear constraints, to which we apply a global lexitions for optimality that lead to efficient algorithms. © 1997 Acacographic optimality criterion. demic PressThe goal of optimal bit allocation traditionally has been to minimize an additive distortion measure, typically meansquared error (MSE), averaged over coding blocks. While INTRODUCTIONthis approach leverages the wealth of tools from optimizaIn any lossy coding system, there is an inherent tradeoff tion theory and operations research, it does not guarantee between the rate of the coded data and the distortion of the constancy in quality that is generally desired from a the reconstructed signal. Often the transmission (storage) video coding system. For example, a video sequence with medium is bandwidth (capacity) limited. The purpose of a constant or near-constant level of distortion is more desirrate control is to allocate bits to coding units and to regulate able than one with lower average distortion but higher the coding rate to meet the bit-rate constraints imposed variability, because human viewers tend to find frequent by the transmission or storage medium while maintaining changes in quality more noticeable and annoying. A long an acceptable level of distortion.video sequence typically contains segments that, even if We consider rate control in the context of the MPEG-1 encoded at a fairly low bit rate, will not contain any disand MPEG-2 video coding standards. In addition to speci-turbing quantization artifacts, so that improving the quality fying a syntax for the encoded bitstream and a mechanism of pictures in those segments is far less important than for decoding it, the MPEG standards define a hypothetical improving the quality of pictures in segments that are more decoder called the video buffering verifier (VBV), which difficult to encode. places quantifiable limits on the variability in bit rate ofTo address these issues, we propose a lexicographic optimality criterion that better expresses the desired constancy the encoded vid...

Constant quality video coding using video content analysis

“…The authors argue that in many applications, the MINMAX criterion is more appropriate than the MINAVE criterion. An extension of the MINMAX approach is the lexicographically optimal approach [10], where the distortion is equalized across the coding units. In [10], the lexicographic criterion is used to find optimal bit allocation under CBR and VBR constraints.…”

“…An extension of the MINMAX approach is the lexicographically optimal approach [10], where the distortion is equalized across the coding units. In [10], the lexicographic criterion is used to find optimal bit allocation under CBR and VBR constraints. The quantizer level is used as the distortion measure and constant quantizer segments are proven to be the optimal solution.…”

Optimal rate control for video transmission over VBR channels based on a hybrid MMAX/MMSE criterion

“…Å for all (9) This new formulation now guarantees that increasing does not lead to TB underflow and any violation of the condition in (2).…”

“…A criterion for minimizing distortion in lexicographical sense (MLEX) has been proposed as a modified MMAX approach to increase overall quality [9]. This criterion is used to find optimal bit allocation under CBR constraints in [9] where quantizer levels are used as a distortion measure and it is shown that the optimal solution is determined by using constant quantizer segments.…”

“…This criterion is used to find optimal bit allocation under CBR constraints in [9] where quantizer levels are used as a distortion measure and it is shown that the optimal solution is determined by using constant quantizer segments. But, in general, if a distortion measure can take any arbitrary values, the proposed algorithm cannot be easily applied.…”