Relative Pixel Value Comparison for Image Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image identification methods fail to accurately determine similarities between images with the same content but different color gradations by comparing pixel values, leading to inaccurate results.
Innovation Solution
An image information acquiring system that calculates relative pixel values and orientations, storing these as information for comparison, allowing for the determination of image similarity by matching these relative values between two images, with a threshold of 80% or higher considered as the same image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pixel values are compared directly to determine image similarity, then the comparison process is simple, but the identification accuracy deteriorates when images have different color gradations
Solution Approach 1:
The patent transforms the comparison parameters from absolute pixel values to relative relationships between adjacent pixels. For each pixel, it calculates the difference with neighboring pixels and represents this as a directional pattern (e.g., up, down, left, right, or combinations). This parameter transformation makes the comparison invariant to uniform color shifts while preserving structural information, thereby improving identification accuracy without requiring complex processing.
Solution Approach 2:
The patent divides the image into individual pixels and further segments each pixel's information into directional relationships with its neighbors. By analyzing the pattern of differences in multiple directions (up, down, left, right, and diagonals), it creates a segmented representation that captures local structural characteristics. This segmentation allows the system to compare images based on their structural patterns rather than absolute color values, resolving the contradiction between simplicity and accuracy.
2Reliability
If absolute pixel values are used for comparison, then the data processing is straightforward, but the system cannot distinguish between images with same structure but different color schemes
Solution Approach 1:
The patent changes the parameter representation from absolute pixel intensities to relative differences between adjacent pixels. For each pixel position, it calculates the difference with neighbors in various directions and encodes this as a directional pattern. This transformation preserves the structural relationships and relative intensity variations while eliminating the influence of overall color shifts, thereby maintaining reliability in similarity determination without losing essential structural information.
Solution Approach 2:
The patent introduces an intermediary representation layer between the raw pixel values and the final comparison. This intermediary consists of the calculated directional difference patterns that serve as a mediator to translate absolute color values into structural relationship descriptors. This intermediary layer filters out the irrelevant color gradient information while preserving the meaningful structural patterns, enabling reliable similarity determination.
Data Source
AI summary
An image information acquiring system obtains a group of relative values to describe the image. A pixel value of each pixel in the image is acquired. Ratios between the pixel value of each pixel and a pixel value of each other pixel which is adjacent to the pixel are obtained for obtaining the group of relative values. The group of relative values is used for comparing two images or reproducing another image.