Convolution Kernel Coefficient Merging for Image Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing systems are inefficient in performing convolution operations due to the high number of multiplication operations and storage requirements for kernel coefficients, which hampers their performance in tasks like edge detection and demosaicing.
Innovation Solution
The method involves applying a convolution operation using a kernel with coefficients that share common values, where the sum of corresponding image data values is multiplied by the common coefficient value, reducing the number of multiplication operations and storage needed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional convolution operation is used with individual coefficients, then processing accuracy is maintained, but the number of multiplication operations and storage requirements increase significantly
Solution Approach 1:
The patent merges multiple coefficient values into common coefficient values that are shared across multiple positions in the kernel. Instead of treating each coefficient independently, the invention groups coefficients with identical values and performs a single multiplication operation for all positions sharing that common coefficient, thereby reducing the total number of multipliers required while maintaining convolution accuracy
Solution Approach 2:
The patent changes the parameter representation from individual coefficients to common coefficient values. By identifying and utilizing the fact that multiple coefficients may have identical values, the invention transforms the convolution operation to use these shared parameters, reducing computational complexity and hardware requirements without sacrificing processing capability
2Productivity
If traditional convolution operation is used with individual coefficients, then complete image processing capability is maintained, but storage space for kernel coefficients increases
Solution Approach 1:
The patent merges multiple coefficient storage locations into a single storage location for common coefficient values. By grouping coefficients that have identical values and storing them once with a reference to all positions where they are used, the invention significantly reduces the total storage space required for kernel coefficients while maintaining complete image processing capability
Solution Approach 2:
The patent makes the stored common coefficient values universal by allowing a single stored value to serve multiple positions in the kernel simultaneously. This multi-functional approach to coefficient storage enables the system to maintain complete processing capability with reduced storage requirements, as each common coefficient value serves multiple purposes in the convolution operation
Data Source
AI summary
An image processing method including obtaining image data. The image data includes a plurality of image data values. The image processing method also includes processing the image data, thereby generating output data. Processing the image data includes applying a convolution operation to the plurality of image data values using a kernel including a plurality of coefficients. Applying the convolution operation includes obtaining a sum of image data values of the plurality of image data values that correspond respectively to coefficients of the plurality of coefficients that each have a common coefficient value. Applying the convolution operation also includes multiplying the sum by the common coefficient value.