Image Feature Compression Using ECC Check Matrices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data compression methods require excessive computation and result in variations that decrease search hit rates due to the large number of multiplications needed for dimensional compression, especially when dealing with high-dimensional data.
Innovation Solution
The use of a check matrix, specifically a check matrix of an error correcting code, such as a binary or LDPC code, for dimensional compression reduces the computational load by minimizing the number of additions and multiplications required, maintaining the relative distance relationships between feature quantities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional random projection matrix is used for dimensional compression, then data compression is achieved, but the computational load increases exponentially and search hit rate decreases
Solution Approach 1:
The patent changes the parameter of the projection matrix from conventional random values to values derived from error correcting codes (ECC). Specifically, it uses parity check matrices from ECC schemes such as Hamming codes, BCH codes, or LDPC codes, where matrix elements are determined by systematic rules rather than random selection. This parameter change reduces computational complexity while maintaining or improving search accuracy, as ECC-based matrices preserve distance relationships more effectively than random matrices.
Solution Approach 2:
The patent employs error correcting code structures (originally designed for error detection and correction in data transmission) as a template for constructing the projection matrix. By copying the mathematical structure of ECC parity check matrices and applying it to dimensional compression, the invention leverages the proven distance-preserving properties of ECC to improve search hit rates while reducing computational burden through the structured nature of these codes.
2Speed
If dimensional compression is performed with high compression ratio, then data processing speed increases, but the amount of computation required for compression increases exponentially
Solution Approach 1:
The patent transforms the projection matrix construction from random value generation to systematic ECC-based value determination. This parameter change enables the use of structured algorithms with lower computational complexity, reducing the energy required for compression operations while maintaining high compression ratios. The systematic structure of ECC matrices allows for more efficient multiplication operations compared to random matrices.
3Ease of manufacture
If random projection matrix elements are used, then simple matrix construction is achieved, but variations in compression results occur and hit rate decreases
Solution Approach 1:
The patent replaces random parameter selection with deterministic parameter generation based on error correcting code theory. The projection matrix elements are systematically determined by ECC parity check relationships, ensuring consistent and reproducible compression results. This systematic approach eliminates the variations inherent in random matrix construction while maintaining construction simplicity through the use of well-established ECC algorithms.
Data Source
AI summary
A dimensional compression unit 13 multiplies a feature quantity extracted from an image by a check matrix of an error correcting code which consists of binary elements each of which is 0 or 1, to compress the image.