Edge Face Recognition via Pixel Padding and Adaptive Boost
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Facial recognition technologies face challenges when a human face is partially or entirely located at the edge of an image, as not all facial features may be included, making it difficult to apply ratio-based recognition techniques effectively.
Innovation Solution
A method is introduced where specified numeric pixels are added to the edge area of an image to generate an enhanced digital image, allowing for successful human face recognition even when the face is partially outside the image boundaries, using a pre-trained adaptive boost human face classifier to identify facial features within sub-images extracted from the enhanced image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If facial recognition relies on ratio-based techniques analyzing facial features, then recognition accuracy is improved when all facial features are visible, but recognition fails when the face is located at the edge area and not all features are included
Solution Approach 1:
The patent applies preliminary action by pre-training an adaptive boost human face classifier before actual recognition. This pre-trained classifier is then used to identify and locate facial features even when they are partially visible at image edges, allowing the system to adapt to edge-area faces without requiring complete facial feature visibility
Solution Approach 2:
The patent changes the approach from fixed ratio-based measurements to adaptive parameter detection using a pre-trained classifier. By transforming the recognition method from relying on predetermined facial feature ratios to using learned parameters from training data, the system can accurately recognize faces regardless of their position in the image
2Adaptability or versatility
If the image data is extended to include edge areas with added pixels, then the ability to recognize faces at image edges is improved, but the complexity of image processing increases
Solution Approach 1:
The patent applies segmentation by dividing the image processing into distinct stages: first adding specified numeric pixels to edge areas, then extracting sub-images, and finally applying the pre-trained classifier. This segmented approach manages complexity by breaking down the overall task into manageable steps rather than attempting to solve the entire problem at once
Data Source
AI summary
A facial recognition solution is disclosed that includes adding a specified numeric pixels to an edge area of a digital image to acquire an enhanced digital image and then performing a facial recognition process on the enhanced digital image to determine a human face from the digital image.