Face De-identification via Virtual Feature Synthesis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing technologies that blur or mosaic face areas in images to prevent facial recognition often degrade image quality and disrupt user concentration.
Innovation Solution
An image processing method that extracts a first face feature from an input image, generates a second face feature by combining the extracted feature with a virtual face feature, and replaces the original face image with a new face image based on the second face feature, thereby de-identifying the face naturally while maintaining image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If face area is processed in blurred or mosaic manner to prevent facial recognition, then portrait rights and privacy are protected, but image quality is degraded and user concentration is disrupted
Solution Approach 1:
The patent extracts only the essential identifying features of the face (facial landmarks, contours, and key geometric relationships) while removing detailed texture and color information. This extraction approach protects privacy by eliminating recognizable details while preserving the overall face structure to maintain image quality and natural appearance.
Solution Approach 2:
The patent applies different processing qualities to different parts of the face image. Critical identifying regions (eyes, nose, mouth) are heavily processed to protect privacy, while non-critical regions maintain higher quality. This local differentiation ensures effective de-identification while preserving overall image aesthetics and user engagement.
2Reliability
If face area is processed in blurred or mosaic manner to prevent facial recognition, then portrait rights and privacy are protected, but user concentration on image is disrupted
Solution Approach 1:
The patent extracts and preserves the fundamental geometric structure and spatial relationships of facial features while removing identifying details. This maintains the natural appearance and composition of the face in the image, allowing users to remain engaged and concentrated while privacy is protected through selective feature removal.
Solution Approach 2:
The patent transforms facial image parameters by changing the level of detail representation. It converts high-detail texture and color information into low-detail geometric structures, altering the image parameters to protect privacy while maintaining the overall visual composition that sustains user concentration.
3Reliability
If conventional blurring or mosaicing is applied to de-identify faces, then privacy infringement issues are addressed, but the natural appearance of the image is lost
Solution Approach 1:
The patent segments the face into key geometric components (facial landmarks, contours, and structural elements) rather than applying uniform blurring. By preserving the segmented geometric structure while removing detailed information, it maintains the natural shape and appearance of the face while effectively protecting privacy.
Solution Approach 2:
The patent changes the representation parameters of facial features from detailed pixel-based information to simplified geometric descriptions. This parameter transformation preserves the essential shape and structure of facial features, maintaining natural appearance while eliminating identifying details for privacy protection.
Data Source
AI summary
A variety of exemplary embodiments for processing images are disclosed. In an exemplary embodiment, an image processing method includes a step of extracting a first face feature from an input image including a first face image, a step of generating a second face feature by combining the extracted first face feature with a virtual face feature, a step of generating a second face image on the basis of the generated second face feature, and a step of generating an output image by substantially replacing the first face image with the generated second face image in the input image.