Face Identity Verification via Covariance Matrix Entropy Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face recognition technologies struggle to determine whether multiple images contain faces from the same person, especially under varying illumination and occlusion conditions, due to limitations in existing feature detection methods.
Innovation Solution
An image processing method that calculates a covariance matrix from target images to determine an upper information entropy limit, which is then used to differentiate between images of the same person and those of different individuals by comparing it against pre-set threshold values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional face detection characteristics (Haar, LBP, HOG) are used, then face detection speed and basic accuracy are acceptable, but the system cannot determine whether faces in multiple images are from the same person
Solution Approach 1:
The patent transforms the face verification problem from traditional feature matching to statistical parameter analysis. By calculating covariance matrices from face images and deriving information entropy limits, the system changes the analytical parameters from appearance features to statistical distribution characteristics, enabling identity determination through mathematical threshold comparison
Solution Approach 2:
The patent replaces traditional mechanical feature matching approaches with mathematical statistical methods. Instead of manually designing and comparing feature vectors, the system uses automated covariance matrix calculation and information entropy computation to objectively determine face identity, substituting manual feature engineering with automated statistical analysis
2Reliability
If face detection is performed under varying illumination and occlusion conditions, then robustness requirements increase, but existing methods cannot reliably determine face identity consistency
Solution Approach 1:
The patent converts the harmful effects of illumination variation and occlusion into beneficial statistical information. By calculating covariance matrices that capture the distribution characteristics of pixel values, the system transforms environmental variations into measurable statistical parameters, where the information entropy limit becomes a robust indicator of face identity that is invariant to illumination and occlusion conditions
Solution Approach 2:
The patent introduces covariance matrices and information entropy limits as intermediary statistical measures between the raw face images and the identity determination decision. These intermediaries aggregate the effects of illumination and occlusion variations, providing a stable statistical representation that enables reliable identity verification despite environmental interference
Data Source
AI summary
An image processing method, a terminal and a storage medium. The method includes: acquiring N target image, wherein each of the target images contains a face image, and N is an integer greater than one (101); determining a covariance matrix of the N target image (102); determining an upper information entropy limit according to the covariance matrix (103); determining whether the upper information entropy limit is greater than a first pre-set threshold value (104); when the upper information entropy limit is greater than the first pre-set threshold value, determining that the N target images contain face images of different people (105); and when the upper information entropy limit is less than or equal to the first pre-set threshold value, determining that the N target images contain face images of the same person (106).