Image Signature Generation for Multi-Texture Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods fail to effectively classify and search images based on macro-scale textures with multiple patterns, as they are typically designed for images with a single fine-grained texture, leading to inefficiencies in identifying and matching images with coarse-grained or multi-texture images.
Innovation Solution
A method that generates a translation-independent signature for images by overlaying a window to define sub-images, determining luminance values, and combining them into a histogram, which is insensitive to the location of the patterns, allowing for effective comparison and search of images with macro-scale textures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing image processing methods are used, then fine-grained single texture images can be classified and searched effectively, but macro-scale multi-texture images cannot be properly processed
Solution Approach 1:
The image is divided into multiple overlapping patches, each processed independently to extract local texture features. This segmentation allows the method to handle different textures within the same image by treating each region separately, then combining results to form a comprehensive texture signature.
Solution Approach 2:
The method transforms image patches into frequency domain representations using Fourier transforms, changing the parameter space from spatial domain to frequency domain. This parameter transformation enables effective characterization of texture patterns regardless of their scale or orientation in the original image.
2Loss of information
If location-dependent texture analysis is used, then spatial information is preserved, but translation invariance is lost
Solution Approach 1:
The method moves from spatial domain analysis to frequency domain analysis using Fourier transforms. By representing textures in the frequency domain, the method achieves translation invariance because frequency characteristics remain consistent regardless of the patch's position in the image, while still capturing essential spatial texture patterns.
3Device complexity
If statistical models are applied to macro-scale textures, then computation is simplified, but matching accuracy deteriorates
Solution Approach 1:
The method replaces traditional statistical modeling approaches with frequency domain analysis using Fourier transforms. This substitution provides a more robust mathematical framework for characterizing textures, capturing periodic patterns and structural information that statistical models miss, thereby improving matching precision while maintaining computational efficiency.
Data Source
AI summary
A method for processing an image to generate a signature which is characteristic of a pattern within said image. The method comprising receiving an image; overlaying a window at multiple locations on said image to define a plurality of sub-images within said image, with each sub-image each having a plurality of pixels having a luminance level; determining a luminance value for each said sub-image, wherein said luminance value is derived from said luminance levels of said plurality of pixels; and combining said luminance values for each of said sub-images to form said signature. Said combining is such that said signature is independent of the location of each sub-image. A method of creating a database of images using said method of generating signatures is also described.