Adaptive Image Blurring for Privacy and Recognition Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face recognition systems that blur images to protect privacy often compromise collation accuracy due to uniform reduction in image information, making it difficult to identify individuals and maintain high security standards.
Innovation Solution
An information processing system that determines the degree of information reduction based on the attributes of subjects in images, such as clothing colors, and applies adaptive blurring intensities to balance privacy protection with collation accuracy by adjusting the amount of information reduction according to the appearance frequency of attributes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If uniform blurring processing is applied to all images to protect privacy, then privacy protection is improved, but collation accuracy deteriorates
Solution Approach 1:
The patent applies different blurring intensities to different regions of images based on the identifiability of subjects. Faces that are easily identifiable receive stronger blurring, while faces that are harder to identify receive weaker blurring. This local differentiation maintains privacy protection for high-risk subjects while preserving collation accuracy for low-risk subjects.
Solution Approach 2:
The patent dynamically adjusts the blurring intensity parameter based on the appearance frequency and identifiability characteristics of subjects in images. By changing the degree of information reduction according to subject attributes, the system optimizes the balance between privacy protection and collation accuracy for different types of subjects.
2Object-affected harmful factors
If high blurring intensity is applied to maximize privacy protection, then privacy protection is improved, but collation accuracy deteriorates
Solution Approach 1:
The system evaluates the identifiability of each subject and applies localized blurring strategies. High blurring intensity is applied only to faces with high identifiability risk, while lower intensity is applied to faces with low identifiability risk. This selective approach ensures privacy protection where needed while maintaining reliability for accurate collation.
Solution Approach 2:
Instead of applying excessive blurring to all images, the system applies partial blurring only to the extent necessary for privacy protection. By calculating the minimum required blurring intensity based on subject attributes, the system avoids excessive information loss while still achieving adequate privacy protection.
3Measurement precision
If low blurring intensity is applied to maintain collation accuracy, then collation accuracy is improved, but privacy protection deteriorates
Solution Approach 1:
The patent implements differentiated blurring strategies based on subject characteristics. Faces with high identifiability risk receive strong blurring for privacy protection, while faces with low identifiability risk receive weak or no blurring to maintain collation accuracy. This local quality differentiation resolves the contradiction by applying appropriate measures only where necessary.
Solution Approach 2:
The system dynamically adjusts the blurring intensity parameter according to the appearance frequency and identifiability of subjects. By changing the degree of information reduction based on subject attributes, the system can maintain privacy protection for vulnerable subjects while preserving collation accuracy for subjects that do not require enhanced protection.
Data Source
AI summary
An information processing device, an information processing system, an information processing method, and a program capable of appropriately reducing the amount of information of an image are provided. An information processing device includes an image acquisitor, an information amount reduction degree determiner, and an information amount reducer. The image acquisitor acquires an image acquired by imaging an actual space. The information amount reduction degree determiner determines a degree of reduction of an amount of information on the basis of an attribute of a subject shown in the image. The information amount reducer generates information reduced data acquired by reducing at least a part of the amount of information of the image in accordance with the degree of reduction of the amount of information.