Depth Map Generation Using Patch Masking for Occlusion Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating depth maps, such as block matching, struggle with occlusions and thin objects, and are computationally intensive, leading to inefficiencies in processing speed and accuracy.
Innovation Solution
The proposed method selects reference pixels in one image and associates them with candidate pixels in another, applying masking operations to select subsets of pixels with similar values, calculating a cost function based on these subsets to determine disparity and depth values, thereby improving handling of occlusions and reducing processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If block matching method is used for depth map estimation, then depth map can be generated, but occlusions and thin objects are not handled well
Solution Approach 1:
The patent divides the image into multiple overlapping patches instead of processing the entire image at once. Each patch is processed independently to determine depth values, which improves handling of occlusions and thin objects by focusing on local regions. The overlapping nature ensures continuity and reduces errors at patch boundaries.
Solution Approach 2:
The patent applies different processing strategies to different regions of the image based on local characteristics. By analyzing local patch properties and applying adaptive matching, the system achieves better depth estimation accuracy for specific regions while maintaining overall reliability in handling occlusions and thin structures.
2Productivity
If traditional depth map estimation methods are used, then depth values can be calculated, but processing time is excessive
Solution Approach 1:
By segmenting the image into smaller patches, the computational complexity is reduced from processing the entire image to processing individual patches. This allows for faster parallel processing and significantly reduces the overall computation time while maintaining depth map quality.
Solution Approach 2:
The patent processes only necessary regions of the image in detail while using coarser processing for other areas. This selective approach reduces the total number of computations required, improving processing speed without sacrificing critical depth information in important regions.
Data Source
AI summary
A depth map is generated from at least a first and a second image. A plurality of reference pixels are selected in the first image. A cost function is used to associate each reference pixel with a respective pixel in the second image. A masking operation is used to identify a subset of pixels in a block of pixels surrounding a reference pixel and the cost function is based on the identified subset of pixels. A disparity between each reference pixel and the respective pixel in said second image is determined, and a depth value is determined for each reference pixel as a function of the respective disparity. A depth map is generated based on the determined depth values.