Hybrid Stereo Imaging via Downscaled SGM Hints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional stereoscopic analysis algorithms for depth estimation in applications like autonomous driving require significant processing capacity, memory, and bandwidth, which is not feasible for vehicles with limited resources, and existing solutions are not efficient enough for real-time operations.
Innovation Solution
A hybrid approach that uses a scaled version of the semi-global matching (SGM) algorithm, where SGM is run on downscaled images to reduce performance requirements, and the results are used as hints for accurate disparity estimation in a dedicated high-performance optical flow/stereo disparity estimation module.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional stereoscopic analysis algorithms are used for depth estimation, then measurement precision is improved, but device complexity and processing requirements increase significantly
Solution Approach 1:
The patent introduces an intermediary component - a downscaler that reduces image resolution before processing. This intermediary transforms the input images from high resolution to lower resolution, enabling conventional stereoscopic algorithms to run on resource-constrained devices while maintaining acceptable depth estimation accuracy. The downscaler acts as a mediator between the high-precision algorithm requirements and the limited processing capacity of embedded systems.
Solution Approach 2:
The patent changes the resolution parameter of the input images by applying a downscaling operation. By reducing the spatial resolution of images before feeding them to the stereoscopic analysis algorithm, the computational complexity is significantly reduced while the depth estimation functionality is preserved. This parameter transformation allows the system to operate within limited processing budgets.
2Measurement precision
If high-resolution images are processed for accurate disparity estimation, then measurement precision is improved, but bandwidth requirements increase significantly
Solution Approach 1:
The patent extracts only the essential information needed for disparity estimation by downsampling images before processing. Instead of transmitting and processing all high-resolution pixel data, the system extracts the critical depth-related information at a lower resolution level. This extraction approach significantly reduces bandwidth consumption while maintaining the capability to perform accurate disparity estimation for depth mapping applications.
3Measurement precision
If conventional stereo algorithms are implemented in real-time applications, then measurement precision is improved, but productivity decreases due to processing time
Solution Approach 1:
The patent performs a preliminary action by downsampling images before they enter the main stereoscopic processing pipeline. This pre-processing step reduces the computational load on subsequent processing stages, enabling real-time operation. By preparing the data in advance at a lower resolution, the system can perform depth estimation faster while maintaining acceptable accuracy for real-time applications such as autonomous driving and robotic navigation.
Data Source
AI summary
A hybrid matching approach can be used for computer vision that balances accuracy with speed and resource consumption. Stereoscopic image data can be rectified and downsampled, then analyzed using a semi-global matching (SGM) process. The use of downsampled images greatly reduces time and bandwidth requirements, while providing high accuracy disparity results. These disparity results can be provided as external hints to a fast module that can perform a robust matching process in the time needed for applications such as real time navigation. The external hints can be used, along with potentially other hints, to define a search space for use by the fast module, which can result in higher quality disparity results obtained within specified timing constraints and with limited resources. The disparity results can be used to determine distances to various objects, as may be important for vehicle navigation or robotic task performance.