Rice Parameter Derivation Using Markov Decision Process
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The computation of the Rice parameter for transform domain coefficients in versatile video coding (VVC) is computationally costly due to a substantial number of memory access and arithmetic operations, especially when entropy coding residual pixels in the transform domain.
Innovation Solution
A decoding method that uses a Markov decision process (MDP) to reduce computational complexity by determining the Rice parameter based on a current state and next state, replacing the more intensive computation of locSumAbs and Rice parameter lookup with less complex operations, and efficiently decoding transform domain coefficients.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the Rice parameter is derived using the conventional method with locSumAbs computation and lookup tables, then the coding accuracy is maintained, but the computational complexity and memory access requirements increase substantially
Solution Approach 1:
The patent changes the parameter representation from using locSumAbs (sum of absolute values) to using a simplified state variable that tracks the cumulative sum directly. This parameter transformation eliminates the need for complex lookup tables and reduces computational operations while maintaining the ability to derive the Rice parameter accurately for Golomb-Rice coding of transform coefficients.
Solution Approach 2:
The patent extracts and removes the computationally intensive locSumAbs computation and lookup table operations from the decoding process. By replacing these with a simpler state-based approach, the invention eliminates unnecessary computational steps while preserving the essential functionality of Rice parameter derivation.
2Measurement precision
If the Rice parameter derivation uses extensive memory access and arithmetic operations, then the coding precision is maintained, but the decoding speed and processing efficiency decrease
Solution Approach 1:
The invention transforms the Rice parameter derivation by changing from a lookup-table-based approach (requiring memory access) to a computational approach using a simple state variable. This parameter change maintains precision in Rice parameter determination while dramatically improving decoding speed by eliminating memory access bottlenecks.
Solution Approach 2:
The patent replaces the mechanical lookup table system with a computational state-based system. Instead of accessing memory tables to find Rice parameters, the decoder uses arithmetic operations on a state variable, substituting a memory-intensive mechanical system with a faster computational approach.
3Reliability
If the conventional Golomb-Rice coding method is used with full Rice parameter computation, then the coding completeness is ensured, but the number of arithmetic operations and processing time increase
Solution Approach 1:
The patent performs preliminary computation by maintaining a state variable that accumulates the sum of absolute transform coefficients as decoding progresses. This preliminary action allows the Rice parameter to be derived immediately without requiring post-computation lookup tables, ensuring coding completeness while reducing processing time.
Solution Approach 2:
The decoding process serves itself by using the state variable (cumulative sum) that is naturally maintained during decoding to directly compute the Rice parameter. This self-service approach eliminates the need for separate lookup table operations, ensuring complete and accurate coding while minimizing additional processing time.
Data Source
AI summary
A video encoding method includes obtaining a video bitstream; encoding the video bitstream based on a residual coding scheme and signaling information; and outputting the encoded video bitstream. Encoding the video bitstream includes: obtaining residual information associated with the video bitstream, wherein the residual information includes at least a remainder; obtaining a transform coefficient based on a current state, a Rice parameter, and a next state; and encoding the residual information based on the transform coefficient.