Transform Coefficient Scanning for High-Frequency Video Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compression systems, such as H.264, inefficiently handle high-frequency data due to Huffman tables tuned for low-frequency residues, resulting in higher bit-rates and poor compression efficiency when dealing with complex motion patterns like rotation and zoom.
Innovation Solution
The proposed system introduces flexibility in interpreting syntax elements like 'total_zeros' and 'run_before' by allowing a novel zigzag reverse scan order for high-frequency data, enabling efficient representation and reducing bit-rate through an optional mode flag indicating the counting mode used during quantization and entropy encoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If Huffman tables are tuned for low-frequency residues, then compression efficiency for low-frequency data is improved, but compression efficiency for high-frequency data deteriorates
Solution Approach 1:
The patent implements dynamic selection of scanning orders (zigzag forward or reverse) based on the frequency characteristics of the transform coefficients. The system adapts the scanning approach to match the data characteristics, using reverse zigzag scanning for high-frequency data and forward zigzag scanning for low-frequency data, thereby resolving the contradiction between optimized low-frequency compression and high-frequency data handling capability
Solution Approach 2:
The patent changes the scanning parameter (scanning order direction) based on the frequency content of the data. By detecting whether the transform coefficients represent high-frequency or low-frequency characteristics, the system switches between different scanning modes, allowing the compression system to maintain high efficiency across different frequency ranges rather than being fixed for low-frequency optimization
2Quantity of substance
If conventional zigzag forward scan order is used, then low-frequency data is efficiently compressed, but high-frequency data requires more bits and achieves poor compression
Solution Approach 1:
The patent applies reverse zigzag scanning for high-frequency data, which scans the transform coefficients in the opposite direction of the conventional forward zigzag scan. This inversion of the scanning order allows the entropy coding stage to more efficiently encode high-frequency coefficients by grouping zeros and non-zero values in a manner that reduces the number of bits required, thereby improving compression efficiency for high-frequency content
3Quantity of substance
If lossy compression is employed, then data compactness is improved, but information loss increases
Solution Approach 1:
The patent applies different scanning and encoding strategies to different frequency regions of the transform coefficients. By treating high-frequency and low-frequency coefficients differently through selective scanning orders, the system optimizes compression for each region's characteristics while minimizing information loss in perceptually important areas, thereby achieving better balance between compactness and information preservation
Data Source
AI summary
A system for compressing digital data by representing a portion of it predictionally and transformationally as a block of transform coefficients, then quantizing that block selectively into a set of encoding symbols based on an indication whether the transform coefficients represent the portion as having a particular characteristic, and then by encoding the set of encoding symbols into a data bit stream. In particular, frequency may be used as the characteristic of the digital data in many applications.