Weighted Color Codeword Encoding for Gradient Image Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression methods struggle to efficiently handle image blocks with slowly varying transitions between two or more colors, particularly in 3D graphics rendering, which requires high memory bandwidth and processing power, leading to suboptimal quality and increased power consumption.
Innovation Solution
The method involves decomposing images into blocks of 2m × 2n image elements, assigning color weights to each element, and determining two or more color codewords based on these weights to represent the original colors through weighted combinations, allowing for effective handling of smoothly varying transitions within and across blocks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of substance
If traditional image compression methods (S3TC, PACKMAN, iPACKMAN) are used, then compression ratio is improved, but image quality deteriorates in blocks with slowly varying color transitions
Solution Approach 1:
The patent changes the parameter representation from fixed palette indices to weighted combinations of reference colors. By representing each pixel color as a linear combination of two reference colors with weights summing to 1, the method achieves continuous color variation within blocks while maintaining compression. This parameter transformation allows accurate representation of slowly varying color transitions that traditional discrete palette methods cannot handle effectively.
2Manufacturing precision
If more colors are used per block to represent gradual transitions, then image quality is improved, but compression ratio deteriorates
Solution Approach 1:
The patent applies local quality by using different representation strategies for different parts of the block. Two reference colors are stored for each block, and each pixel within the block uses a weighted combination of these local reference colors. This allows each block to have its own localized color space optimized for representing gradual transitions, achieving high image quality without increasing global compression overhead.
3Manufacturing precision
If complex decompression algorithms are used to handle color transitions, then image quality is improved, but processing power requirements increase
Solution Approach 1:
The patent replaces complex mechanical decompression algorithms with simple mathematical operations. Instead of using sophisticated prediction or interpolation algorithms, the method uses basic linear combination operations with pre-stored reference colors. The decompression process involves only multiplication and addition of color components, which are computationally inexpensive operations that can be efficiently implemented in hardware or software.
4Loss of substance
If higher compression ratios are achieved, then memory bandwidth is reduced, but rendering complexity increases
Solution Approach 1:
The patent performs preliminary action by pre-calculating and storing two reference colors for each block during compression. These reference colors are chosen to best represent the color range within the block. During rendering, the system only needs to retrieve these pre-computed reference colors and apply simple weighted combinations, avoiding the need for complex real-time color calculations. This preliminary preparation significantly reduces rendering complexity while maintaining high compression ratios.
Data Source
Figure 1~2
Figure 3~5
Figure 4~6
AI summary
In an image-encoding scheme, an input image is decomposed into several image blocks (600) comprising multiple image elements (610). The image blocks (600) are encoded into encoded block representations (700). In this encoding, color weights are assigned to the image elements (610) in the block (600) based on their relative positions in the block (600). At least two color codeword (710, 720, 730, 740) are determined, at least partly based on the color weights. These codewords (710, 720, 730, 740) are representations of at least two color values. The original colors of the image elements (610) are represented by color representations derivable from combinations of the at least two color values weighted by the assigned color weights.