Parallax Calculation Using Hierarchical Block Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing position measurement devices using stereo cameras face limitations in accuracy due to occlusion issues and insufficient matching window sizes, which affect the precision of distance calculations.
Innovation Solution
The proposed arithmetic device improves distance calculation accuracy by using a hierarchical structure for image processing, where a first parallax search unit calculates parallax in smaller blocks based on both parent and adjacent blocks' parallax candidates, and a candidate calculation unit converts distance information into parallax candidates for detailed distance image generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a conventional correlation calculation method is used with fixed matching window size, then the calculation process is simple, but the measurement precision is insufficient due to occlusion issues and inability to adapt to different regions
Solution Approach 1:
The patent divides the image into multiple blocks and further segments each block into sub-blocks for correlation calculation. This hierarchical segmentation allows different matching window sizes to be applied to different regions, improving measurement precision by adapting to local features while managing complexity through systematic division of the calculation process
Solution Approach 2:
The patent implements dynamic adjustment of matching window sizes based on block characteristics. The calculation process adapts to different regions by selecting appropriate window sizes, transforming the static fixed-window approach into a dynamic system that responds to local image features, thereby improving precision without requiring overly complex manual configuration
2Measurement precision
If a small matching window size is used to capture fine details, then the edge sharpness is improved, but the measurement precision deteriorates due to insufficient sampling and increased noise sensitivity
Solution Approach 1:
Each block is divided into multiple sub-blocks, allowing the system to use small effective matching windows within sub-blocks for edge sharpness while aggregating results from multiple sub-blocks to improve noise tolerance. This segmentation enables simultaneous achievement of fine detail capture and robust noise handling
Solution Approach 2:
The patent combines results from multiple sub-blocks within each block to produce the final distance measurement. This merging process improves reliability by averaging out noise while preserving the edge sharpness achieved through small sub-block matching windows, effectively combining the advantages of both small and large window approaches
3Reliability
If a large matching window size is used to reduce noise, then the noise tolerance is improved, but the measurement precision deteriorates due to loss of edge detail and reduced spatial resolution
Solution Approach 1:
The patent segments each block into multiple sub-blocks, allowing large effective coverage for noise reduction while maintaining small local matching windows that preserve edge details. This segmentation enables the system to achieve both noise tolerance and edge sharpness simultaneously
Solution Approach 2:
The patent introduces a hierarchical dimension to the matching process, operating at multiple levels (blocks and sub-blocks). This dimensional approach allows large windows to be used at the block level for noise reduction while small windows are used at the sub-block level for edge preservation, effectively resolving the contradiction through multi-scale processing
4Measurement precision
If the entire image is processed at full resolution, then the measurement precision is improved, but the calculation time increases significantly
Solution Approach 1:
The patent divides the image into multiple blocks and processes each block independently with further sub-block segmentation. This segmentation enables parallel processing of different regions, reducing overall calculation time while maintaining full resolution processing for accurate distance measurements in each local area
Solution Approach 2:
The patent applies correlation calculation selectively to sub-blocks rather than processing the entire image uniformly. By focusing computational resources on relevant regions and using hierarchical processing, the system achieves sufficient measurement precision without the excessive calculation time that would result from processing all pixels at full resolution
Data Source
Figure 1
Figure 2(a)~2(c)
Figure 3
AI summary
An arithmetic device includes a storage unit that stores a coarse distance image indicating a parallax in units of first blocks, a first parallax search unit that uses a standard image that is an image captured by a first camera and a reference image that is an image captured by a second camera to calculate, in units of second blocks having a size equal to or smaller than that of the first block, a parallax in the standard image based on a parallax candidate that is a candidate for a parallax for each of the second blocks, and generate a detailed distance image that is a distance image in detail, and a candidate calculation unit that calculates the parallax candidate based on a parallax of a parent block that is a region having a size of the first block including the second block for which the first parallax search unit calculates a parallax, a parallax of an adjacent block that is a region having the size of the first block adjacent to the parent block, and a search condition.