Hybrid Depth Regularization for Noisy Stereo Depth Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing depth estimation methods in computer stereo vision are prone to noise and errors due to factors like textureless regions, non-Lambertian surfaces, occlusions, and spatially varying lens characteristics, which affect the accuracy of depth maps, especially in real-time applications such as augmented reality.
Innovation Solution
A hybrid depth regularization process that categorizes scene regions and applies region-specific techniques, including smoothing filters and edge-preserving filters, to improve depth map quality by combining multiple depth estimates and confidence maps, while handling occlusions and textureless areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional depth estimation methods are used, then processing speed is maintained, but depth map accuracy deteriorates due to noise and errors in textureless regions, non-Lambertian surfaces, occlusions, and spatially varying lens characteristics
Solution Approach 1:
The patent segments the depth estimation process into multiple independent stages: a first depth estimation process that handles textureless regions and occluded areas, and a second depth estimation process that processes textured regions. This segmentation allows each process to be optimized for its specific challenges, improving overall accuracy without requiring a single complex unified system.
Solution Approach 2:
The patent applies different depth estimation approaches to different regions of the image based on their characteristics. Textureless regions, non-Lambertian surfaces, and occluded areas are processed by one method, while textured regions are processed by another method. This local differentiation enables optimized processing for each region type, resolving the contradiction between accuracy and complexity.
2Measurement precision
If multiple depth estimation processes are combined to improve accuracy, then depth map quality improves, but computational load increases
Solution Approach 1:
The computational task is segmented into two separate depth estimation processes that operate on different image regions. The first process handles textureless and occluded regions, while the second process handles textured regions. This segmentation reduces the computational burden on each individual process compared to a single comprehensive process, while still achieving high overall accuracy through combination.
Solution Approach 2:
Instead of applying a single comprehensive depth estimation process to the entire image, the patent applies partial processes tailored to specific regions. Each process performs only the necessary computations for its designated regions, avoiding unnecessary computational overhead in regions handled by the other process, thus reducing total computational load while maintaining accuracy.
Data Source
AI summary
Systems and methods for hybrid depth regularization in accordance with various embodiments of the invention are disclosed. In one embodiment of the invention, a depth sensing system comprises a plurality of cameras; a processor; and a memory containing an image processing application. The image processing application may direct the processor to obtain image data for a plurality of images from multiple viewpoints, the image data comprising a reference image and at least one alternate view image; generate a raw depth map using a first depth estimation process, and a confidence map; and generate a regularized depth map. The regularized depth map may be generated by computing a secondary depth map using a second different depth estimation process; and computing a composite depth map by selecting depth estimates from the raw depth map and the secondary depth map based on the confidence map.