Stereo Vision Disparity Map Generation via Multi-Resolution Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In stereo vision, high-resolution images lead to increased computation and time complexity for accurate disparity estimation, making it challenging to achieve efficient depth information generation.
Innovation Solution
The method involves capturing images with dual sensors, performing a shrinking process to generate shrunk images, calculating shrunk disparity maps, magnifying these maps, and refining disparity values based on reliability to produce a refined disparity map for accurate depth estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution images are used for disparity estimation, then measurement precision is improved, but computation complexity and time consumption increase
Solution Approach 1:
The image processing is divided into multiple resolution levels. The method first processes a down-sampled (lower resolution) version of the image to obtain an initial disparity map, then uses this as guidance for processing the full-resolution image. This segmentation by resolution levels reduces the overall computational complexity while maintaining accuracy.
Solution Approach 2:
The method performs preliminary disparity estimation on the down-sampled image before processing the full-resolution image. This preliminary disparity map serves as a guide for the subsequent fine-tuning process, allowing the algorithm to start with a rough estimate and progressively refine it, rather than attempting to compute the complete high-resolution disparity map from scratch.
2Measurement precision
If high-resolution images are used for disparity estimation, then measurement precision is improved, but time consumption increases
Solution Approach 1:
The computation is segmented into two time-efficient stages: first processing the smaller down-sampled image to get an initial disparity map quickly, then using this as a starting point for the full-resolution processing. This segmentation reduces total time consumption compared to processing only the full-resolution image.
Solution Approach 2:
The down-sampled disparity estimation serves as a preliminary action that provides a head start for the final high-resolution estimation. By having this preliminary result ready, the subsequent fine-tuning process requires fewer iterations and less time to achieve the same level of accuracy.
3Device complexity
If down-sampled images are used for disparity estimation, then computation complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The method transitions from a single-resolution approach to a multi-resolution approach, adding the dimension of resolution scaling. By working in multiple resolution dimensions (down-sampled and full-resolution), the method achieves both computational efficiency and high precision, overcoming the limitations of using only down-sampled images.
Solution Approach 2:
The down-sampled disparity map serves as a preliminary guide that directs the subsequent full-resolution processing. This preliminary action ensures that the computationally expensive full-resolution processing is focused on refining specific areas, thereby achieving high precision without the full computational burden of processing the entire high-resolution image independently.
Data Source
AI summary
A method, an image processing device, and a system for generating a disparity map are proposed. The method includes the following steps. Images of a specific scene are captured by two image sensors to generate two original images. A shrinking process is performed on the two original images to generate two shrunk images. A shrunk disparity map is generated by using the two shrunk images. A magnifying process is performed on the shrunk disparity map to generate a magnified disparity map. Whether each magnified disparity value of the magnified disparity map is reliable is determined so as to accordingly generate a refined disparity map.