Sparse Orthonormal Transforms for Image Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image and video compression technologies, such as those using the Discrete Cosine Transform (DCT), struggle with efficiently compressing blocks containing strong edges, leading to increased bitrate due to many non-zero coefficients, and prior solutions like wavelets lack block-based granularity and directionality preservation.
Innovation Solution
The use of sparse orthonormal transforms that classify blocks based on directional structure, optimizing transforms and coefficients iteratively to minimize rate distortion, allowing for the selection of the best transform for each block and reducing the number of coefficients needed for representation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If DCT is used for image and video compression, then compression is successful for general blocks, but bitrate increases when blocks contain strong edges not aligned in horizontal or vertical direction due to many non-zero coefficients
Solution Approach 1:
The patent segments the transform domain by providing multiple predefined transform matrices corresponding to different edge orientations. Instead of using a single DCT for all blocks, the system divides the transform space into multiple segments, each optimized for specific directional characteristics, thereby reducing the number of non-zero coefficients for edge-containing blocks
Solution Approach 2:
The patent dynamically selects the appropriate transform matrix based on the detected edge orientation in each block. The transform choice is not fixed but adapts to the local characteristics of each block, allowing the system to optimize the transform operation for each specific case and minimize non-zero coefficients
2Adaptability or versatility
If wavelet transforms are used, then decomposition is achieved, but block-based granularity and directionality preservation are lost which are required for state-of-the-art video and image codecs
Solution Approach 1:
The patent applies segmentation by dividing the image or video into discrete blocks and applying different transform matrices to each block based on its specific characteristics. This maintains the block-based structure required by modern codecs while achieving adaptive decomposition similar to wavelets
Solution Approach 2:
The patent implements local quality by assigning different transform properties to different blocks based on their local edge characteristics. Each block receives a transform optimized for its specific directional content, preserving local directionality while maintaining the overall block-based structure
3Productivity
If prediction from previously decoded portions is used, then compression is achieved, but the solution is not suitable for video compression where residual signal is not correlated beyond motion compensation block boundaries
Solution Approach 1:
The patent changes the parameter being predicted from pixel values or block values to edge orientation parameters. By predicting and encoding the orientation of edges rather than the full block content, the system achieves compression that is applicable to both image and video contexts where residual correlation patterns differ
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
A method and apparatus is disclosed herein for performing compression with sparse orthonormal transforms. In one embodiment, the method comprises receiving a block of data; classifying the block of data based on directional structure; selecting one of a plurality of available directional, orthonormal transforms to apply to the block based on results of classifying the block; and applying the selected transform to the block to generate a plurality of coefficients, thereby producing a compressed version of the block.