Stereo Matching System with Adaptive Window and Backward Tracer
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional 3D imaging systems with fixed window sizes for stereo matching result in either noisy or blurred depth maps, and are calculation-intensive, occupying large circuit areas and causing latency.
Innovation Solution
A system using adaptive window sizes for stereo matching, incorporating a stereo matching unit and a backward tracer to generate depth maps with reduced circuit area and latency, and a compensation unit to fill hole regions, with window sizes dynamically adjusted based on texture density using edge detectors and texture density units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If a fixed small window size is used for stereo matching, then the circuit area and latency are reduced, but the depth map becomes noisy
Solution Approach 1:
The patent implements dynamic window size adjustment where the window size is no longer fixed but adapts based on local image characteristics. The system calculates texture density for different regions and dynamically selects window sizes (e.g., 3x3, 5x5, 7x7, or 9x9) to match the local complexity, thereby resolving the contradiction between using small windows for efficiency and large windows for quality.
Solution Approach 2:
The patent applies different window sizes to different regions of the image based on local texture density. High-frequency regions (edges, textures) use smaller windows to preserve detail and avoid noise, while low-frequency regions (smooth areas) use larger windows to improve matching accuracy. This local adaptation resolves the contradiction by optimizing window size for each specific region rather than using a uniform size.
2Manufacturing precision
If a fixed large window size is used for stereo matching, then the depth map quality is improved, but the circuit area and latency increase
Solution Approach 1:
The system dynamically adjusts window size based on local image characteristics, using larger windows only where necessary (in smooth regions) and smaller windows in complex regions. This dynamic adaptation maintains depth map quality where needed while reducing overall processing time and computational load compared to using a uniformly large window size.
Solution Approach 2:
By applying larger window sizes only to low-frequency regions where they provide benefit and using smaller windows in high-frequency regions, the system optimizes the balance between quality and processing efficiency. This local differentiation avoids the unnecessary computational overhead of using large windows throughout the entire image.
3Adaptability or versatility
If conventional stereo matching is applied to multi-view applications, then depth maps can be generated for multiple views, but the calculation intensity and circuit area requirements increase significantly
Solution Approach 1:
The patent merges the stereo matching operations for multiple views by reusing the depth map from one view as input for generating depth maps in other views. Instead of performing independent stereo matching for each view pair, the system combines operations through the backward tracer that propagates depth information across multiple views, significantly reducing computational redundancy and circuit area requirements.
Solution Approach 2:
The backward tracer component serves multiple functions: it derives depth maps for multiple views from a single reference depth map, performs hole filling compensation, and adapts to different view configurations. This multi-functional component enables the system to handle various multi-view applications without requiring separate dedicated hardware for each view, thereby reducing overall device complexity.
Data Source
AI summary
A system of image stereo matching includes at least one stereo matching unit (SMU) each receives a first view and a second view of a view pair, according to which the SMU generates a first depth map for the first view. The system also includes a backward tracer operable to receive the first depth map, according to which a second depth map for the second view is derived.