Human Figure Region Extraction Using Face Detection and Iterative Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing human figure region extraction methods in images are inaccurate due to inappropriate segmentation, where regions generated through segmentation can be too small or large, leading to incorrect judgment of human figure regions and inclusion of background areas.
Innovation Solution
A method that detects a face or facial part in an image, determines an estimated region based on the face's position, extracts the human figure region, and iteratively extends and updates the region to include nearby outer regions if the human figure is partially present in the outline periphery, using skin color information and evaluation values for pixel classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If automatic region segmentation is applied to extract human figure regions, then extraction automation is improved, but extraction accuracy deteriorates due to inappropriate segmentation degree causing regions to be too small or too large
Solution Approach 1:
The patent applies preliminary action by first detecting the face region before extracting the human figure region. The face detection step prepares the system by identifying a reliable anchor point (the face) that is then used to determine the estimated human figure region, improving both automation and accuracy
Solution Approach 2:
The patent divides the extraction process into multiple segmentation stages: face detection, estimated region determination, and human figure region extraction. This multi-stage segmentation approach allows each stage to focus on specific tasks, improving overall extraction accuracy while maintaining automation
2Speed
If the estimated region is determined based only on face position, then processing speed is improved, but extraction completeness deteriorates when the human figure extends beyond the initial estimated region
Solution Approach 1:
The patent applies feedback by iteratively checking whether the extracted human figure region touches the outline periphery of the estimated region. If it does, the system feeds back by extending the estimated region and repeating the extraction process, ensuring complete human figure extraction while maintaining efficiency through conditional iteration
Data Source
AI summary
Upon extraction of a human figure region in an image, a face or facial part is detected in the image, and an estimated region which is estimated to include the human figure region is determined from position information of the detected face or facial part. The human figure region is extracted in the estimated region. Judgment is made as to whether at least a portion of the human figure region exists in an outline periphery region of the estimated region, and the estimated region is extended and updated so as to include a near outer region near the human figure region in the outline periphery region and outside the estimated region, in the case where a result of the judgment is affirmative. The human figure region is extracted in the extended and updated estimated region.