Image Content Structure Analysis via Autocorrelation and Entropy Metrics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods struggle to effectively discern the structure of images, particularly periodic patterns, due to their complexity and variability in scale, frequency, and location, which can be masked by normalized phase plane correlation and are not adequately revealed by simple Fourier transforms.
Innovation Solution
A method and apparatus that correlates image data, computes image statistics, estimates eccentricity and entropy, and generates a report on image content characteristics using autocorrelation, radon transforms, and watershed algorithms to interpret and classify image structure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If normalized phase plane correlation is used to analyze periodic patterns, then the analysis covers the full range of periodicities from featureless to chaotic content, but the characteristics in the content structure are masked and the analysis becomes not useful
Solution Approach 1:
The patent segments the analysis by introducing a new metric space that separates periodicity detection from content structure analysis. Instead of using a single normalized phase plane correlation surface, the patent divides the analysis into distinct components: one for detecting periodicity presence and another for analyzing content structure characteristics, thereby avoiding the masking effect while maintaining comprehensive coverage.
Solution Approach 2:
The patent introduces an intermediary metric transformation that converts the normalized phase plane correlation data into a new representation space. This intermediary step allows the system to preserve the full range of periodicity information while transforming the data into a form where content structure characteristics remain visible and analyzable without being masked.
2Device complexity
If simple Fourier transform is used for periodic pattern detection, then the method is computationally simple, but it cannot effectively reveal the essential properties of periodic patterns with varying scales and frequencies
Solution Approach 1:
The patent transitions from the traditional frequency domain analysis of Fourier transform to a new metric space that incorporates both spatial and frequency characteristics. By introducing additional dimensional information through the phase plane correlation metric and subsequent transformations, the system achieves comprehensive periodic pattern detection without the computational complexity of traditional multi-method approaches.
3Quantity of substance
If multiple periodic patterns are spatially co-located and super-positioned, then the image contains rich structural information, but the normalized phase plane correlation masks the characteristics of individual patterns
Solution Approach 1:
The patent applies local quality analysis by examining the metric space at different locations and scales independently. Instead of applying a global normalization that masks local characteristics, the system analyzes each region's periodic patterns with appropriate local metrics, preserving the unique characteristics of each co-located pattern while accounting for their super-positioning effects.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and apparatus of processing image data comprises correlating received image data. Image statistics are computed based upon the correlated image and eccentricity is estimated based upon the computed image statistics. An entropy metric of the correlated received image data is determined. An interpretation based upon the image statistics, estimated eccentricity, and entropy metric is performed and a report including the content of the processed image data is generated.