Face Recognition via Resized Image Mapping for Wide-Angle Lenses
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Digital image processing apparatuses with wide-angle lenses face challenges in face recognition due to smaller face sizes in images, making smile and blink detection difficult, which affects the image photographing success rate.
Innovation Solution
A digital signal processing unit generates images of different sizes, maps face detection information from a resized image onto a live-view image, and performs smile and blink recognition by comparing coordinates and characteristics with stored database information, ensuring accurate detection even for small face sizes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of moving object
If a wide-angle lens is used to provide a wider viewing angle, then the viewing angle is improved, but the face size in the image becomes smaller which deteriorates face recognition ratio
Solution Approach 1:
The patent divides the image processing into multiple stages: first detecting faces in a resized version of the image (where faces are larger and easier to detect), then mapping the detected face coordinates back to the original live-view image. This segmentation allows the system to leverage the wide viewing angle capability while maintaining accurate face recognition through coordinate transformation.
Solution Approach 2:
The patent introduces a resized image as an intermediary representation. Instead of directly analyzing faces in the original live-view image where they appear small, the system creates a resized version where faces are enlarged for easier detection, then uses coordinate mapping to transfer the detection results back to the original image space. This intermediary enables both wide-angle coverage and accurate face recognition.
2Productivity
If the image is resized to a smaller size to match the live-view image dimensions, then the processing efficiency is improved, but the face size becomes too small for accurate smile and blink detection
Solution Approach 1:
The patent segments the detection process into two parts: first performing face detection on a resized image (where faces are sufficiently large for accurate analysis), then mapping the results to the original image. This allows the system to process images efficiently at smaller dimensions while maintaining detection accuracy through the coordinate transformation step.
Solution Approach 2:
The patent performs preliminary face detection on a resized version of the image before mapping to the original image. By preparing the detection results in advance on the resized image where faces are larger and easier to analyze, the system avoids the difficulty of detecting smiles and blinks on extremely small faces in the original image, while still maintaining processing efficiency.
Data Source
AI summary
An apparatus for improving a face recognition ratio includes a digital signal processing unit which generates a first image having a size different from that of a live-view image and, when the size of a face detected from the first image is smaller than a reference size, maps information about the face detected from the first image on the live-view image and performs smile and/or blink recognition.