Face Image Composite Generation for Masked Subjects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies struggle to accurately generate composite images of faces that are partially hidden, such as those obscured by masks, leading to bland and unattractive photographs in situations where individuals are hesitant to remove their masks.
Innovation Solution
An information processing apparatus and method that includes units for acquiring images, detecting face areas, estimating hidden shield areas, estimating expressions, generating estimated expression images, and creating composite images to reveal the obscured face areas based on estimated expressions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If mask-wearing is maintained for health safety, then protection against disease transmission is improved, but face visibility and photograph attractiveness deteriorate
Solution Approach 1:
The system creates a copy of the person's face by estimating and generating a synthetic facial region that mimics the appearance and expression of the actual face. This copy is then combined with the masked image to produce a composite photograph that shows the face without the mask, thereby resolving the contradiction between maintaining mask protection and achieving face visibility in photographs.
2Measurement precision
If the entire face area is used for identification, then recognition accuracy is improved, but privacy protection deteriorates when the face is fully visible
Solution Approach 1:
The system applies different processing qualities to different regions of the face. The shielded region (where the mask covers the face) is estimated and generated with synthetic facial features for identification purposes, while the unshielded region (visible areas like eyes and forehead) maintains its original appearance for privacy protection. This allows accurate face recognition while preserving privacy in visible areas.
3Object-affected harmful factors
If shielded regions are simply removed or blanked out, then privacy is protected, but image completeness and attractiveness deteriorate
Solution Approach 1:
The system performs preliminary estimation of the shielded facial region before final image composition. By predicting what the concealed facial features would look like based on visible portions and expression analysis, the system prepares the synthetic facial data in advance, allowing for complete and attractive face reconstruction in the final composite image while maintaining privacy during the processing stages.
Data Source
AI summary
An information processing apparatus includes: an acquisition unit that acquires information about a person including at least an image of the person; a detection unit that detects a face area including a face of the person from the image; an estimation unit that estimates a hidden shield area in a case where at least a part of the face area is hidden; an expression estimation unit that estimates an expression of the person on the basis of the information about the person; an estimated expression image generation unit that generates an estimated expression image of an area corresponding to the shield area, in accordance with the expression estimated by the expression estimation unit; and a composite image generation unit that generates a composite image on the basis of the image and the estimated expression image.