Skin Image Anonymization Using Autoencoder Feature Masking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Images of skin used in research and development often contain privacy-sensitive information such as tattoos, birthmarks, or skin lesions, which can compromise user privacy and security.
Innovation Solution
A computer-implemented method using an autoencoder model to process skin image data, identifying and altering target regions with skin features by adjusting code values based on a distribution of non-target region values to generate altered image data that masks or obscures these features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If skin images are used for research and development, then research utility is improved, but user privacy is compromised due to identifying skin features
Solution Approach 1:
The patent extracts and removes identifying skin features (tattoos, birthmarks, lesions) from skin images while preserving the underlying skin structure and texture. This is achieved through segmentation of the skin image into different layers and selective modification of the identifying features, allowing research to proceed with anonymized images that no longer contain personally identifiable information
Solution Approach 2:
The patent applies different processing qualities to different regions of the skin image. Identifying features are heavily processed or removed, while surrounding normal skin areas are preserved with high fidelity. This local differentiation allows the image to maintain research value for skin analysis while eliminating privacy risks associated with unique identifying markers
2Object-affected harmful factors
If identifying skin features are removed from images, then privacy is protected, but image utility for research may be reduced
Solution Approach 1:
The patent segments the skin image into multiple layers or regions, separating identifying features from the underlying skin structure. By processing only specific segments containing identifying features while leaving other segments intact, the method protects privacy without sacrificing the research value of the skin's structural information, texture, and other scientifically relevant characteristics
3Object-affected harmful factors
If skin images are processed to remove identifying features, then privacy security is improved, but image realism may be reduced
Solution Approach 1:
The patent creates realistic replacements for removed identifying features by synthesizing alternative skin features that match the surrounding skin characteristics. Instead of simply leaving blank areas or applying obvious masks, the system generates synthetic skin features (such as alternative moles, pores, or texture patterns) that are statistically consistent with normal skin variations, thereby maintaining image realism while ensuring privacy protection
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Proposed concepts thus aim to provide methods, and systems pertaining to processing image data of skin. In particular, embodiments aim to provide a method for altering a target region of the image data that depicts a skin feature of a subject.