Image Similarity Determination Using Pixel Pair Sequences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods for determining similarities between images are inaccurate when one image is rotated, and feature comparison methods fail to capture detailed similarities.
Innovation Solution
An image processing device with a similarity determination system that converts images to grayscale, matches pixels using an exhaustive or backtracking method to create a pixel pair sequence, and calculates a similarity value based on grayscale differences, ensuring accurate similarity assessment regardless of image orientation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If pixel comparison method is used to determine image similarity, then the method is simple to implement, but the accuracy deteriorates when one image is rotated to a different orientation
Solution Approach 1:
The patent applies preliminary action by extracting feature points and establishing feature point correspondence between images before performing similarity determination. This preprocessing step creates a robust foundation that remains valid even when images are rotated, allowing the subsequent similarity calculation to be orientation-independent and thus resolving the accuracy deterioration problem while maintaining implementation simplicity.
2Measurement precision
If feature comparison method is used to extract basic features such as grayscale and texture features, then the method can reflect overall similarities between images, but it fails to capture detailed similarities between images
Solution Approach 1:
The patent applies segmentation by dividing the image comparison task into two distinct levels: feature point level for detailed local similarity and pixel level for overall global similarity. By segmenting the similarity determination into these two components, the method captures both detailed similarities (through feature point correspondence) and overall similarities (through pixel comparison), thereby resolving the information loss problem while maintaining comprehensive similarity assessment.
3Measurement precision
If exhaustive search method is used to match pixels between images, then the accuracy of pixel matching is improved, but the computational complexity and time consumption increase
Solution Approach 1:
The patent applies preliminary action by performing feature point matching and establishing correspondence before conducting pixel-level exhaustive search. This preliminary feature point alignment constrains the search space and provides a reference framework that guides the subsequent pixel matching process, thereby maintaining high pixel matching accuracy while reducing overall computational time compared to pure exhaustive search on all pixels.
Data Source
AI summary
A computerized method determines similarities between a first image and a second image. The first image is converted into a first grayscale image, and the second image is converted into a second grayscale image, where a number of pixels of the second image is not less than a number of pixels of the first image. Each pixel of the first grayscale image is matched with a pixel of the second grayscale image to create a pixel pair sequence. A similarity value indicating the similarities between the first image and the second image is calculated according to the grayscale value of each pixel in the pixel pair sequence S.