Distance Map Generation Using Segmentation and Propagation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques for generating depth maps from stereo images often face a trade-off between accuracy and computational resources, leading to sparse disparity maps and inefficiencies in real-time applications, especially in low-textured scenes or with sparse projected light patterns.
Innovation Solution
The method involves generating an initial distance map using block or code matching algorithms and a segmentation map using segmentation algorithms, where image elements are grouped into segments based on similar color or grayscale values, allowing for the assignment of updated distance values that reduce sparseness and improve accuracy, enabling real-time or near-real-time processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If global optimization algorithms are used to produce full disparity maps, then measurement precision and reliability are improved, but computational time and power consumption increase significantly
Solution Approach 1:
The patent divides the image into multiple regions or segments, and processes each segment separately using simplified algorithms. This segmentation allows the system to achieve full disparity maps without requiring computationally expensive global optimization across the entire image, thus reducing processing time while maintaining acceptable accuracy in each local region.
Solution Approach 2:
The patent applies partial action by using simplified matching algorithms that do not perform complete global optimization but instead use local or regional optimization strategies. This partial approach sacrifices some global consistency but achieves sufficient accuracy for real-time applications while dramatically reducing computational requirements.
2Productivity
If simplified matching algorithms are used, then computational speed is improved, but measurement precision and completeness deteriorate due to incorrect or missing disparity values
Solution Approach 1:
The patent merges multiple processing stages: initial simplified matching to generate sparse disparity values, segmentation to identify regions, and propagation to fill in missing values. By combining these approaches, the system achieves both speed (from simplified algorithms) and completeness (from propagation techniques).
Solution Approach 2:
The patent introduces an intermediary propagation step that acts as a mediator between the simplified matching algorithm and the final disparity map. This intermediary process takes the sparse results from fast algorithms and propagates them through the segmented regions to produce complete, accurate disparity maps without requiring expensive global optimization.
3Device complexity
If simplified matching algorithms are used, then device complexity is reduced, but the resulting disparity map becomes sparsely populated with missing values
Solution Approach 1:
The patent segments the image into multiple regions and processes each segment independently. This segmentation allows the use of simple algorithms within each region while using propagation techniques to ensure complete coverage, thereby maintaining low complexity while avoiding information loss.
Solution Approach 2:
The patent performs preliminary segmentation and region identification before the matching process. This preliminary action allows the system to prepare propagation paths and region boundaries in advance, enabling the simple matching algorithms to work efficiently while ensuring that no areas are left uncovered in the final disparity map.
Data Source
AI summary
Techniques are described for generating a distance map (e.g., a map of disparity, depth or other distance values) for image elements (e.g., pixels) of an image capture device. The distance map is generated based on an initial distance map (obtained, e.g., using a block or code matching algorithm) and a segmentation map (obtained using a segmentation algorithm). In some instances, the resulting distance map can be less sparse than the initial distance map, can contain more accurate distance values, and can be sufficiently fast for real-time or near real-time applications. The resulting distance map can be converted, for example, to a color-coded distance map of a scene that is presented on a display device.