Pattern Image Detection via Block Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional pattern detection methods for template images in target images are inefficient due to the need for individual comparison of absolute positions and pixel values, leading to increased computation time and noise errors, especially when dealing with large template images.
Innovation Solution
A pattern image detection method that divides both the template and target images into regions of equal size, calculates the sum of squared brightness values for each region, and performs a normalized cross-correlation calculation to determine similarity, allowing for efficient comparison and reduced computation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all absolute positions and pixel values of pixels are compared individually, then detection accuracy is maintained, but detection computation time is increased
Solution Approach 1:
The patent divides the template image into multiple blocks (e.g., 8x8 pixel blocks) and processes each block separately. This segmentation allows the system to compare pixel values in groups rather than individually, reducing the total number of comparisons needed while maintaining detection accuracy through block-level correlation calculations.
Solution Approach 2:
The patent implements a two-stage comparison process: first performing a coarse comparison using block-level statistics (mean, variance) to quickly eliminate non-matching regions, then performing detailed pixel-level comparison only in regions that pass the initial filter. This partial action approach maintains accuracy for promising candidates while saving time by avoiding exhaustive comparison of all pixels.
2Measurement precision
If template image size is increased to capture more pattern information, then detection accuracy is improved, but computation time is further increased
Solution Approach 1:
The patent divides large template images into multiple smaller blocks that can be processed independently and in parallel. This segmentation reduces the computational burden of processing large images while maintaining overall detection accuracy by aggregating results from all blocks through normalized cross-correlation.
Solution Approach 2:
The patent performs preliminary filtering using block-level statistical features (mean intensity, variance) before conducting full pixel-level correlation. This partial action approach allows the system to handle large template images by first identifying promising regions through quick statistical comparison, then applying computationally intensive correlation only where necessary.
3Reliability
If individual pixel comparison is performed, then noise errors are increased, but no preprocessing is applied
Solution Approach 1:
The patent combines multiple pixels into blocks and computes aggregate statistics (mean, variance) for each block. This merging approach reduces the impact of individual pixel noise by averaging effects across multiple pixels, thereby improving reliability while managing processing complexity through efficient statistical computations.
Solution Approach 2:
The patent performs preliminary calculations of block-level statistical features (mean intensity, variance) before conducting the actual correlation comparison. This preliminary action prepares the data in advance, reducing noise effects through averaging and enabling faster subsequent comparison operations.
Data Source
AI summary
Provided are a method and apparatus for detecting a pattern image, and more particularly, a pattern image detection method and a pattern image detection apparatus for effectively detecting a pattern of a template image from a target image. The pattern image detection method and the pattern image detection apparatus can provide an effect of quickly and accurately detecting the pattern of the template image from the target image while reducing the amount of computation for detecting the pattern of the template image.