Image Processing System Depth Accuracy via Phase Gradient Cost Volume
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing systems face challenges in determining depth information from a single image effectively, particularly in capturing detailed disparity and gradient information, which affects image quality and clarity.
Innovation Solution
An image processing system comprising an image sensor and an image processing device that generates data groups based on phase and brightness information, determines gradient values, and calculates a cost volume by weighted summing these groups to obtain depth information, utilizing threshold values adapted to pixel values and human visual perception principles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If depth information is determined from a single image using conventional methods, then the processing speed is fast, but the measurement precision of depth information is insufficient
Solution Approach 1:
The patent segments the image processing into distinct components: generating multiple data groups from pixel values, calculating gradient values for each group, applying threshold values to determine gradient information, and computing cost volumes. This segmentation allows systematic extraction of depth information while maintaining processing efficiency through structured operations.
Solution Approach 2:
The patent transitions from 2D image pixel values to 3D depth information by generating multiple data groups representing different disparity ranges and calculating cost volumes that incorporate gradient information. This dimensional transformation enables accurate depth determination from single-image 2D data.
2Measurement precision
If multiple data groups are generated and processed with gradient information, then the depth information accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent applies threshold values to gradient values on a local basis for each data group and disparity range, rather than using global processing. This allows selective computation focused on relevant regions and disparity ranges, reducing unnecessary computational operations while maintaining depth accuracy.
Solution Approach 2:
The patent generates multiple data groups covering different disparity ranges and processes gradient information selectively for each group. By focusing computational effort on partial regions (specific disparity ranges and gradient directions) rather than processing all possible combinations, the method achieves accurate depth information with reduced overall computational load.
Data Source
AI summary
An image processing device comprises a data group generator, a gradient value manager, and a cost volume manager. The data group generator is configured to receive pixel values of an image that include phase and brightness information, and determine data groups indicating disparity of the image based on target pixel values, where the data groups correspond to a range that is determined according to the phase information. The gradient value manager is configured to determine gradient values of a region corresponding to the target pixel values and determine gradient information. The gradient information is determined by applying threshold values to the gradient values, where the threshold values are determined according to the target pixel values. The cost volume manager is configured to determine a cost volume by weighted summing the data groups based on the gradient information.