Lensless Camera Mask Processing for Privacy-Preserving Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lensless camera technologies face challenges in ensuring privacy protection, maintaining image quality, and optimizing mask patterns for recognition processing, while also requiring unnecessary image reconstruction and additional processing resources.
Innovation Solution
An image processing apparatus and method that utilizes a mask to modulate light, an image sensor to capture the modulated image, and a signal processing unit to perform signal processing based on the mask pattern, optimizing the mask pattern for privacy and reducing unnecessary image reconstruction and processing resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a mask is divided into multiple sub-areas with individual mask designs and band pass filters, then image quality of reconstructed image is improved, but device complexity and manufacturing cost increase
Solution Approach 1:
The mask is divided into multiple sub-areas, with each sub-area having an individually optimized mask design and band pass filter tailored to specific wavelength ranges, enabling high image quality reconstruction without requiring complex additional components
Solution Approach 2:
Different portions of the mask are assigned different optical characteristics (mask patterns and wavelength filters) optimized for specific functions, allowing simultaneous optimization of multiple image quality parameters across different spectral ranges
2Measurement precision
If a light shielding wall is provided between sub-areas, then boundary portion image quality is improved, but manufacturing difficulty and cost increase
Solution Approach 1:
The light shielding function at sub-area boundaries is merged with the mask structure itself, eliminating the need for separate light shielding walls and simplifying manufacturing while maintaining image quality
3Measurement precision
If mask pattern is optimized to reduce diffraction effect, then image quality is improved, but optimization for recognition processing is compromised
Solution Approach 1:
The mask pattern parameters are optimized specifically for recognition processing tasks rather than general imaging, accepting increased diffraction effects in exchange for superior recognition accuracy and task-specific performance
4Measurement precision
If image reconstruction is performed before recognition processing, then recognition accuracy is improved, but processing resources and power consumption increase
Solution Approach 1:
The essential recognition information is extracted directly from the modulated image captured by the sensor, eliminating the need for full image reconstruction and reducing processing resources while maintaining recognition accuracy
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Achieves highly accurate image recognition with privacy protection by simplifying the apparatus configuration and reducing processing load, while enabling reconstruction of images at various distances without the need for a lens.
Implementation Method 1
a mask configured to modulate and transmit incident light
Implementation Method 2
an image sensor configured to capture a modulated image based on the incident light modulated by the mask
Data Source
AI summary
The present disclosure relates to an image processing apparatus, an image processing method, and a program capable of realizing highly accurate image recognition processing while taking privacy into consideration with a simple configuration. A mask of a lensless camera is convoluted with a weight of a DNN first layer used in image recognition processing using a DNN, and a reconstructed image becomes a processing result of the DNN first layer on the basis of a captured image captured by an image sensor. Accordingly, the restored image becomes the processing result of the DNN first layer and becomes an image difficult for a person to visually recognize as a scene or an object, making it possible to protect the privacy of the image. The present disclosure can be applied to an image recognition apparatus using a lensless camera.