Stereo Matching Intermediate Data Reduction via Block Representative Values
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Semi-global matching (SGM) methods for stereo matching face challenges due to the large size of intermediate data required for processing, which increases memory and bandwidth demands in image processing systems.
Innovation Solution
An image processing apparatus with a first arithmetic unit calculating a first cost function, a second arithmetic unit calculating a second cost function, and a search unit that reduces data by calculating representative values for pixel blocks instead of individual pixels, allowing for efficient disparity value search within a defined search range.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If intermediate data for each pixel is stored and transferred between processing steps, then stereo matching accuracy is maintained, but memory capacity and bandwidth requirements increase significantly
Solution Approach 1:
The patent divides the image processing into multiple steps (cost calculation, intermediate data generation, disparity search) and segments the data processing by grouping pixels into blocks. Instead of processing and storing data for each individual pixel, the system processes pixel blocks together, reducing the overall data volume while maintaining matching accuracy through systematic segmentation of both processing steps and spatial data organization
Solution Approach 2:
The patent introduces a new dimension of processing by calculating representative values for pixel blocks rather than individual pixels. This dimensional shift from pixel-level to block-level processing reduces data volume significantly while the representative values preserve the essential information needed for accurate disparity search, effectively trading spatial resolution for data efficiency
2Quantity of substance
If representative values for pixel blocks are calculated instead of individual pixels, then data storage and transfer volume is reduced, but processing complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-calculating representative values for pixel blocks before the disparity search step. The first arithmetic unit computes cost functions for all pixels in a block, then the second arithmetic unit aggregates these into representative values that are stored and transferred. This preliminary aggregation reduces the data volume that needs to be handled in subsequent processing steps, offsetting the initial computational effort through reduced memory and bandwidth requirements
Solution Approach 2:
The patent introduces representative values as an intermediary data structure between the cost calculation step and the disparity search step. Instead of directly processing raw pixel data through the entire pipeline, the representative values serve as a compressed intermediate representation that preserves essential matching information while reducing data volume, acting as a mediator that simplifies subsequent processing operations
3Loss of information
If large intermediate data is transferred between processing steps, then complete pixel information is preserved, but bandwidth requirements and manufacturing costs increase
Solution Approach 1:
The patent changes the parameter of data representation from individual pixel values to block representative values. By transforming the data from a fine-grained pixel-level representation to a coarser block-level representation, the system reduces the amount of information that needs to be stored and transferred. This parameter change maintains sufficient information completeness for disparity search while significantly reducing bandwidth requirements and associated manufacturing costs for memory and interconnect infrastructure
Data Source
AI summary
According to one embodiment, an image processing apparatus is provided with a first arithmetic unit, a second arithmetic unit, a third arithmetic unit, and a search unit. The first arithmetic unit calculates a first cost function. The first cost function represents a relation between a disparity value and a cost in a search range of a second image. The second arithmetic unit calculates a second cost function. The second cost function is a sum of first cost functions regarding a plurality of paths. The third arithmetic unit calculates a representative value of intermediate data. The intermediate data is an addition result of first cost functions regarding a first path group among the plurality of paths.