Ratio-Based Haar Feature Calculation for Illumination Invariance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face-detection algorithms struggle with illumination variance, requiring memory-intensive and time-consuming statistical variance calculations to normalize Haar-like feature values, which hinders real-time face detection performance.
Innovation Solution
The approach calculates Haar-like feature values as ratios of average pixel intensities within rectangular regions, eliminating the need for statistical variance computation and integral image generation, thereby reducing processing time and memory usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If statistical variance calculations are used to normalize Haar-like feature values, then illumination invariance is improved, but processing time and memory usage increase
Solution Approach 1:
The patent changes the normalization parameter from statistical variance to simple average intensity. Instead of calculating variance which requires multiple passes and memory storage, the method uses average intensity values that can be computed in a single pass, thereby maintaining illumination invariance while reducing processing time
Solution Approach 2:
The patent extracts only the essential normalization information (average intensity) from the full statistical analysis (variance). By taking out only what is necessary for illumination normalization and discarding the computationally expensive variance calculation, the method achieves the same goal with reduced computational overhead
2Productivity
If integral image generation is used for fast feature calculation, then feature computation speed is improved, but memory consumption increases
Solution Approach 1:
The patent performs preliminary calculation of average intensity values directly from the original image during the detection process itself, rather than pre-generating an integral image. This eliminates the need for additional memory storage while still enabling fast feature computation through efficient averaging algorithms
3Measurement precision
If statistical variance computation is performed for each image region, then normalization accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent simplifies the normalization parameter from statistical variance (which measures spread) to average intensity (which measures central tendency). This parameter change reduces computational complexity from O(n) variance calculation to O(1) averaging, while maintaining sufficient normalization accuracy for illumination invariance
Solution Approach 2:
The patent replaces the expensive, complex variance calculation with a cheap, simple averaging operation. The simple average intensity value serves as a disposable, easy-to-compute normalization factor that achieves the same practical purpose without the computational burden
Data Source
AI summary
Faces in images are quickly detected with minimal memory resource usage. Instead of calculating a Haar-like feature value by subtracting the average pixel intensity value in one rectangular region from the average pixel intensity value in another, adjacent rectangular region, a face-detection system calculates that Haar-like feature value by dividing the average pixel intensity value in one rectangular region by the average pixel intensity value in another adjacent rectangular region. Thus, each Haar-like value is calculated as a ratio of average pixel intensity values rather than as a difference between such average pixel intensity values. The feature values are calculated using this ratio-based technique both during the machine-learning procedure, in which the numerical ranges for features in known face-containing images are learned based on labeled training data, and during the classifier-applying procedure, in which an unlabeled image's feature values are calculated and compared to the previously machine-learned numerical ranges.