Video Coding Significant Coefficient Location Encoding Modes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding methods are inefficient in recording the locations of significant coefficients due to sparse distribution, leading to high data rates and costly encoding operations, particularly in transform-based image compression.
Innovation Solution
The method employs multiple encoding modes, including sample-by-sample, point coordinates, quartation, and sixteen-partition modes, to adaptively determine the minimum number of bits required for encoding significant coefficients, selecting the optimal mode based on coefficient distribution patterns and including mode information in the bit stream for accurate decoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If run-length coding for zeros is used, then the encoding of zero coefficients is simplified, but the locations of significant coefficients still require expensive encoding
Solution Approach 1:
The patent segments the block of coefficients into multiple scan patterns (e.g., diagonal, horizontal, vertical, zigzag) and selectively applies different scanning methods to different regions or contexts. This segmentation allows the system to efficiently handle both zero coefficients and significant coefficients by choosing the appropriate scan pattern for each case, reducing the overall encoding cost for location information.
2Ease of operation
If significant coefficients are encoded sample-by-sample, then the encoding process is simple, but a large number of zeros must be encoded to record significant coefficient locations
Solution Approach 1:
The patent extracts and separately encodes the locations of significant coefficients using compact location encoding schemes (such as coordinate-based encoding or differential encoding) rather than encoding them as part of the sample-by-sample sequence. This extraction approach removes the burden of encoding numerous zero coefficients and focuses encoding resources only on the significant coefficients, dramatically reducing the total bit count.
Solution Approach 2:
The patent transitions from one-dimensional sample-by-sample encoding to two-dimensional spatial location encoding by representing significant coefficient positions as coordinates (row, column) within the block. This dimensional change allows the encoder to compactly represent the location of significant coefficients independent of their sequential position, reducing the number of bits required especially when significant coefficients are sparse.
3Productivity
If quartation processing is used to reach significant coefficients, then the number of coding operations is reduced, but the number of bits to encode increases for certain distribution patterns
Solution Approach 1:
The patent implements dynamic selection among multiple encoding modes (e.g., quartation mode, scan mode, coordinate mode, hybrid mode) based on the actual distribution characteristics of significant coefficients in each block. The encoder analyzes metrics such as the number of significant coefficients, their spatial clustering, and block energy distribution to dynamically choose the most efficient mode, thereby optimizing the balance between coding operations and bit count for each specific case.
Solution Approach 2:
The patent changes encoding parameters adaptively by adjusting the encoding mode, scan pattern, and precision of location representation based on the significance distribution. For example, when significant coefficients are concentrated in specific regions, the patent uses region-adaptive scanning or reduced-precision coordinate encoding, thereby reducing the bit count while maintaining accurate location representation.
4Stability of the object's composition
If a single encoding mode is used for all blocks, then the encoding process is consistent, but it cannot adapt to different significance distributions
Solution Approach 1:
The patent creates a universal encoding framework that incorporates multiple encoding modes and patterns within a single system. This multi-functional encoder can switch between different modes (quartation, scanning, coordinate encoding, hybrid approaches) and adapt to various significance distributions while maintaining a consistent overall architecture and interface. The universal design ensures stability in processing flow while providing versatility in handling different coefficient distribution patterns.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
In known image compression, following quantisation, a very sparse distribution of significant (i.e. non-zero) amplitude coefficients of the transformed image signal may be obtained while most quantised coefficients are zeros. A costly task for a transform-based image compression in terms of resulting overall data rate is to record the locations of such significant coefficients within the coding blocks. In quartation processing, a 'significant square' (containing at least one non-zero amplitude coefficient in the coefficient block) is recursively divided into four smaller squares until single significant coefficients are reached, and the significance statuses of all generated squares are encoded. However, for some distribution patterns encoding the x-y-coordinates of the significant coefficients as binary numbers will lead to less coding cost. According to the invention, at least four different pattern determination or encoding modes are checked, and the encoding side selects the least costly one of these modes and transfers the corresponding mode information to the decoding side for corresponding decoding.