Wide-Angle Image Pose Estimation via Region Dewarping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing augmented reality systems using wide-angle image sources, such as fisheye or 360-degree cameras, face unreliability due to distortion of image features, leading to inaccurate pose estimation and unstable positioning of virtual objects in AR environments.
Innovation Solution
The method involves dewarping regions of wide-angle images, estimating poses for the dewarped regions using Structure from Motion (SfM) techniques, and deriving a pose for the wide-angle image by comparing features between dewarped regions, thereby improving the accuracy and stability of device localization and tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If wide-angle image sources (fisheye or 360-degree cameras) are used to capture AR content, then the field of view and coverage area are improved, but image feature distortion increases leading to unreliable pose estimation
Solution Approach 1:
The patent divides the distorted wide-angle image into multiple undistorted image segments or regions. Each region is processed separately to extract features and estimate pose, avoiding the distortion problems that affect the entire wide-angle image. This segmentation allows accurate pose estimation while maintaining the broad field of view advantage.
Solution Approach 2:
The patent introduces an intermediary undistorted image as a mediator between the distorted wide-angle image and the pose estimation process. The distorted image is first transformed into an undistorted representation, which then serves as the basis for reliable feature extraction and pose calculation, eliminating the direct harmful effect of distortion.
2Device complexity
If wide-angle images are used directly for pose estimation, then the system complexity is reduced, but the reliability of AR content positioning deteriorates
Solution Approach 1:
The patent performs preliminary undistortion of the wide-angle image before pose estimation. By pre-processing the image to remove distortion, the subsequent pose estimation and tracking operations can proceed reliably without requiring complex distortion compensation algorithms during real-time operation, thus maintaining system simplicity while improving reliability.
3Measurement precision
If distortion correction is applied to wide-angle images, then pose estimation accuracy is improved, but processing time and computational complexity increase
Solution Approach 1:
The patent segments the wide-angle image into multiple regions that can be undistorted and processed in parallel. This segmentation reduces the computational burden on each individual region and enables concurrent processing, thereby decreasing overall processing time while maintaining accurate pose estimation for each segment.
Solution Approach 2:
The patent applies undistortion selectively to specific regions or portions of the wide-angle image that are most relevant for pose estimation, rather than processing the entire image at full resolution. This partial action approach reduces computational load and processing time while still achieving sufficient accuracy for the AR application.
Data Source
AI summary
The pose of a wide-angle image is determined by dewarping regions of the wide-angle image, determining estimated poses of the dewarped regions of the wide-angle image and deriving a pose of the wide-angle image from the estimated poses of the of the dewarped regions. The estimated poses of the dewarped regions may be determined by comparing features in the dewarped regions with features in prior dewarped regions from one or more prior wide-angle images, as well as by comparing features in the dewarped regions with features in a point cloud.