Luminance-Based Masking for Privacy-Preserving Foreground Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques for surveillance, such as those described in Japanese Patent Application Laid-Open No. 2016-115214, fail to clearly distinguish between foreground objects in crowded scenes when using single-color masking, making it difficult to recognize individuals while maintaining privacy.
Innovation Solution
An information processing apparatus that extracts foreground objects from images, generates a mask image with colors varying based on luminance levels, and combines it with a predetermined image to create an output image, allowing users to differentiate between individuals while protecting privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single-color image is used as a mask image to protect privacy, then privacy protection is achieved, but the boundary between persons and the situation in the captured image becomes unrecognizable
Solution Approach 1:
The patent applies local quality by assigning different colors to different regions of the mask image based on luminance values. Instead of using a uniform single color, each pixel in the mask image receives a color determined by its luminance characteristics, allowing preservation of spatial information and boundaries while maintaining privacy protection.
Solution Approach 2:
The patent utilizes color changes by mapping luminance values to different colors in the mask image. This color variation based on luminance enables the mask image to convey information about the distribution and boundaries of foreground objects, solving the problem of unrecognizable situations while protecting individual privacy.
2Device complexity
If a single-color mask image is used, then processing simplicity is maintained, but the ability to differentiate between multiple individuals is lost
Solution Approach 1:
The patent implements local quality by making each pixel's color dependent on its local luminance properties. This approach maintains processing simplicity through automated luminance-based color assignment while enabling differentiation between individuals through color variations that reflect their respective luminance characteristics.
Solution Approach 2:
The patent applies parameter changes by transforming the single luminance parameter into multiple color parameters. This transformation enables richer information representation in the mask image, allowing differentiation between individuals based on their luminance profiles while maintaining a systematic and relatively simple processing framework.
3Reliability
If foreground objects are extracted and masked with a single color, then privacy protection is achieved, but the visual distinction between different objects is eliminated
Solution Approach 1:
The patent applies local quality by assigning colors locally based on each pixel's luminance value within the foreground object regions. This preserves the visual distinction between different objects and their boundaries while ensuring that no individual can be identified, as the color assignment is based on luminance rather than identity-specific features.
Solution Approach 2:
The patent utilizes color changes to maintain visual distinction between different foreground objects. By varying colors according to luminance values, the mask image preserves the spatial arrangement and boundaries of multiple individuals, enabling users to understand the situation without compromising privacy through identification.
Data Source
AI summary
An information processing apparatus includes an extraction unit, a generation unit, and a combining unit. The extraction unit extracts a foreground object region from a captured image captured by an image capturing unit. The generation unit generates a mask image corresponding to the foreground object region in the captured image, and having colors different depending on luminance of the foreground object region. The combining unit combines a mask image generated by the generation unit and a predetermined image to generate an output image.