Dual-Camera Depth Estimation Using Event Pixel Interpolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image vision processing methods using Dynamic Vision Sensors (DVS) face challenges in accurately determining depth information for non-event pixel points due to inconsistencies in pixel distribution and illumination, leading to mismatching and noise points, which reduces the overall accuracy of depth information.
Innovation Solution
The method involves determining parallax and depth information for event pixel points and using neighboring event pixel points to calculate the depth information of non-event pixel points, thereby avoiding mismatching and improving accuracy, even in regions with difficult illumination or occlusions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If frame image matching technology is used to determine parallax between pixel points, then depth information can be obtained, but mismatching occurs for non-event pixel points leading to reduced accuracy
Solution Approach 1:
The patent segments the pixel points into event pixel points and non-event pixel points, applying different processing strategies to each segment. Event pixel points are processed using traditional frame image matching, while non-event pixel points are handled through interpolation based on neighboring event pixel points, thereby avoiding mismatching issues for non-event pixels.
Solution Approach 2:
The patent introduces an intermediary approach by using event pixel points as reference mediators to determine depth information for non-event pixel points. Instead of directly matching non-event pixel points between frames (which causes mismatching), the method uses neighboring event pixel points as intermediaries to interpolate and determine the depth of non-event pixels.
2Productivity
If non-event pixel points are included in matching, then more pixel points can be processed, but noise points increase due to mismatching
Solution Approach 1:
The patent segments pixel points into event and non-event categories, processing each segment differently. Non-event pixel points are not directly matched but instead have their depth information interpolated from neighboring event pixel points, maintaining high accuracy while achieving complete pixel coverage.
Solution Approach 2:
The patent creates a copy of depth information from neighboring event pixel points to assign to non-event pixel points. By copying depth values from reliable event pixels to non-event pixels through interpolation, the method achieves complete pixel coverage without introducing mismatching errors.
3Ease of operation
If traditional matching is applied to all pixel points, then uniform processing is simplified, but accuracy deteriorates for non-event pixel points
Solution Approach 1:
The patent divides the processing into two distinct segments: event pixel points processed by traditional matching and non-event pixel points processed by interpolation. This segmentation maintains ease of operation for event pixels while dramatically improving accuracy for non-event pixels.
Solution Approach 2:
The patent applies different processing qualities to different regions of the image based on pixel type. Event pixel points receive full matching processing, while non-event pixel points receive interpolated processing, optimizing the overall system performance by matching local characteristics to appropriate processing methods.
Data Source
AI summary
Exemplary embodiments provide an image vision processing method, device and equipment and relate to: determining parallax and depth information of event pixel points in a dual-camera frame image acquired by Dynamic Vision Sensors; determining multiple neighboring event pixel points of each non-event pixel point in the dual-camera frame image; determining, according to location information of each neighboring event pixel point of each non-event pixel point, depth information of the non-event pixel point; and performing processing according to the depth information of each pixel point in the dual-camera frame image. Since non-event pixel points are not required to participate in the matching of pixel points, even if it is difficult to distinguish between the non-event pixel points or the non-event pixel points are occluded, depth information of the non-event pixel points can be accurately determined according to the location information of neighboring event pixel points.