Arbitrary Shape Pattern Matching via FFT Convolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing pattern matching techniques face high computational costs for arbitrary shape pattern searches and are limited by noise when restricting patterns to rectangular shapes, failing to efficiently match patterns of arbitrary shapes.
Innovation Solution
A computer-implemented method using normalized cross-correlation evaluation through convolution of mask and search images, employing Fast Fourier Transform (FFT) and inverse FFT to reduce calculations and allow for arbitrary shape pattern matching, with normalization to zero mean and unit variance for accurate results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If pattern matching is performed for arbitrary shape patterns, then pattern matching capability is improved, but computational cost increases to n4 calculations
Solution Approach 1:
The patent replaces direct spatial convolution (mechanical approach) with frequency domain convolution using FFT. By transforming images to frequency domain, performing multiplication, and transforming back, the computational complexity is reduced from O(n^4) to O(n^2 log n), resolving the contradiction between arbitrary shape support and computational efficiency.
Solution Approach 2:
The patent changes the domain parameter from spatial domain to frequency domain. This transformation allows the same pattern matching operation to be performed with different computational characteristics, achieving both arbitrary shape support and improved efficiency through the FFT-based approach.
2Productivity
If pattern matching is restricted to rectangular shapes, then computational cost is reduced to n2 log n, but pattern matching accuracy deteriorates due to unwanted noise
Solution Approach 1:
The patent segments the rectangular search space by applying a binary mask that identifies only the relevant arbitrary-shaped pattern regions. This segmentation allows the efficient rectangular search algorithm to be applied only to meaningful regions, eliminating noise from irrelevant areas while maintaining computational efficiency.
3Productivity
If arbitrary patterns are inscribed in rectangles, then computational efficiency is improved, but measurement precision deteriorates due to introduced noise
Solution Approach 1:
The patent introduces a binary mask as an intermediary between the rectangular search framework and the arbitrary pattern. The mask acts as a filter that preserves the computational efficiency of rectangular search while eliminating the noise problem by suppressing responses in regions outside the arbitrary pattern boundaries.
Data Source
AI summary
Among other disclosed subject matter, a computer-implemented method for pattern matching includes receiving a pattern image, a mask image and a search image, the mask image having an arbitrary shape and identifying a portion of the pattern image. The method includes evaluating a normalized cross-correlation equation based on the pattern image, the mask image and the search image, including at least a convolution of the mask image and the search image. The method includes outputting a result of evaluating the normalized cross-correlation equation, the result indicating whether the search image matches the portion of the pattern image.