3D Tight Frame Decomposition for Color Image Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for measuring the similarity of image blocks, particularly in color images, fail to adequately consider the intrinsic correlation between channels, leading to low accuracy in image inpainting and artifact formation.
Innovation Solution
A computer-implemented method using a three-dimensional tight frame to decompose images into sub-bands, calculating Laplacian probabilities, and weighting high-frequency sub-bands to measure similarity between image blocks, incorporating channel information through dot multiplication and statistical calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If steerable filter decomposition is used to measure image similarity, then the decomposition process can be performed, but the intrinsic correlation between channels in color images is ignored
Solution Approach 1:
The patent transforms the traditional two-dimensional image decomposition approach into a three-dimensional decomposition by incorporating the channel dimension. The three-dimensional tight frame decomposes the color image into multiple sub-bands across spatial and channel dimensions, allowing simultaneous capture of spatial features and channel correlations. This dimensional extension enables the method to preserve intrinsic channel relationships while performing feature decomposition.
Solution Approach 2:
The patent employs a composite decomposition framework that integrates multiple decomposition results across different channels and scales. By combining the statistical features from each channel's decomposition results through a unified similarity measurement formula, the method creates a composite representation that captures both individual channel characteristics and their inter-channel correlations, thereby resolving the information loss problem.
2Productivity
If existing two-dimensional image feature extraction methods are used, then image blocks can be matched, but the intrinsic correlation of each channel in color images is not fully considered
Solution Approach 1:
The patent segments the color image into multiple channels and performs three-dimensional tight frame decomposition on each channel separately, obtaining distinct sub-bands for each channel. This segmentation allows the method to capture channel-specific features while the subsequent integration of results from all channels preserves their correlations. The segmentation approach enables efficient parallel processing while maintaining measurement precision.
3Ease of operation
If each channel of color images is treated as independent information, then processing is simplified, but the intrinsic correlation between channels is ignored
Solution Approach 1:
The patent merges the decomposition results from multiple channels by computing statistical features (mean, variance, skewness, kurtosis) for each sub-band and integrating these features across all channels using a unified similarity measurement formula. This merging process maintains the simplicity of independent channel processing while simultaneously capturing the intrinsic correlations between channels through the joint statistical analysis, thereby resolving the contradiction between operational ease and measurement accuracy.
Data Source
AI summary
The present application relates to a method for measuring the similarity of images/image blocks, which comprises: S1: acquiring two three-dimensional airspace images V and W; S2: decomposing the images V and W to obtain a plurality of sub-bands; S3: calculating a Laplacian probability corresponding to each high-frequency sub-band of V and W, weighting the high-frequency sub-hand; S4: marking two image blocks as X and Y, taking out data blocks corresponding to the image blocks X and Y, and calculating the statistics of the data blocks; S5: calculating the similarities of X and Y in each channel of each sub-band according to the statistics of the data blocks; S6: calculating an average value of the similarities of X and Y in each channel of each sub-band, and taking the average value as the similarity between X and Y.