Transform-Skipped Block Coding With Adaptive Context States
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In High Efficiency Video Coding (HEVC), the encoding and decoding of transform skipped blocks are inefficient due to incorrect context probabilities, which disrupt the adaptive context model used for conventional transform blocks, leading to increased bit rates and reduced compression efficiency.
Innovation Solution
Adapting the context model for transform skipped blocks by starting in a different state, allowing the reuse of the context model for both transformed and non-transformed blocks, thereby improving the encoding and decoding efficiency without introducing new context models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the standard context model is used for transform skipped blocks, then the encoding process is simple, but the context probabilities become incorrect leading to increased bit rates
Solution Approach 1:
The patent applies preliminary action by pre-determining the starting state based on the transform skip flag value. Before processing the bitmask, the system checks the flag and sets the appropriate starting state (state A for flag=0, state B for flag=1). This preliminary configuration ensures that the correct context probabilities are used from the beginning of the decoding process, avoiding the need for complex runtime adjustments while maintaining high compression efficiency.
2Loss of information
If different context models are created for transformed and non-transformed blocks, then compression efficiency improves, but device complexity increases
Solution Approach 1:
The patent implements universality by designing a single context model that can handle both transformed and non-transformed blocks through the use of different starting states. Instead of maintaining separate context models, the system uses one versatile model that adapts its behavior based on the transform skip flag. This approach achieves the compression efficiency of specialized models while avoiding the complexity of managing multiple separate models.
Solution Approach 2:
The patent applies local quality by allowing different parts of the context model (specifically different starting states) to have different probability characteristics suited for different block types. State A is optimized for transformed blocks while state B is optimized for non-transformed blocks. This enables the system to apply locally optimized probability distributions without creating entirely separate models for each block type.
Data Source
AI summary
Video encoder and decoder and methods therein for transform skipped encoding and decoding of blocks of pixels. The methods comprises a modified use of an existing context model associated with transform coefficients, for deriving contexts for a bitmask indicating which residual values (i.e. not transform coefficients), in a transform skipped residual block, that are greater than one. The methods are applicable to video coding schemes such as HEVC.