Image Privacy Protection Using Depth Map Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies fail to effectively remove privacy-sensitive data like human faces and license plates from images while maintaining the clarity of non-private features such as buildings and navigation information, especially in real-world applications like car navigation systems.
Innovation Solution
The method involves acquiring an input image and its associated depth map, determining the spatial resolution for each pixel, and applying a transformation to reduce the spatial resolution of pixels with a predefined threshold, ensuring that privacy-sensitive data becomes unrecognizable while preserving the clarity of non-private features, using depth information from 3D point clouds or image processing techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If image processing is applied to remove privacy-sensitive data, then privacy protection is improved, but image quality and clarity are degraded
Solution Approach 1:
The patent applies different processing treatments to different regions of the image based on depth information. Regions containing privacy-sensitive objects (faces, license plates) are blurred, while regions containing important navigation information (buildings, signs, streets) are preserved with original quality. This local differentiation resolves the contradiction by protecting privacy only where necessary while maintaining image quality where needed.
Solution Approach 2:
The image is segmented into different regions based on depth map analysis and object detection. The patent divides the image into privacy-sensitive regions (requiring blurring) and non-sensitive regions (requiring preservation). This segmentation allows selective application of privacy protection measures without degrading the overall image quality unnecessarily.
2Measurement precision
If depth information processing is used to identify privacy-sensitive objects, then privacy protection accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent introduces a depth map as an intermediary data structure that stores depth information for each pixel. This depth map serves as a mediator between the raw image and the privacy processing algorithm, enabling accurate identification of privacy-sensitive objects without requiring complex direct analysis of the image content. The depth information acts as a simplified proxy that guides the blurring process.
Solution Approach 2:
The patent performs preliminary processing to generate a depth map and identify privacy-sensitive regions before applying the actual blurring operation. By pre-processing the image to detect faces, license plates, and other sensitive objects using depth information, the system simplifies the subsequent privacy protection step. This preliminary action separates the complex detection task from the simple blurring task.
Data Source
AI summary
A method of reducing the spatial resolution of images is disclosed. At least one embodiment of the method includes: —acquiring an input image including image parts having a spatial resolution larger than SR pixels/meter; —acquiring a depth map associated with the input image; —determining for each pixel p(x,y) a spatial resolution value by means of the depth map; —processing a region of pixels of the input image for which holds that the spatial resolution value is larger than a predefined threshold corresponding to SR pixels/meter to obtain a corresponding region of pixels having a spatial resolution smaller then or equal to SR pixels/meter in an output image. The method enables to removes privacy information from images by reducing the spatial resolution to a level that the privacy information cannot be recognized in the image anymore.