Image Blur Correction Device Using Depth Map Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image blur correction technologies fail to accurately correct image blur caused by camera shake, especially when the parallax amount between images is small, leading to periods of ineffective processing and control.
Innovation Solution
An image blur correction device that tracks feature points, calculates three-dimensional coordinates, and adjusts the imaging apparatus's position attitude using a combination of feature point tracking information, shake sensors, and inertial sensors to drive a shake corrector, ensuring accurate blur correction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature point tracking is performed using SFM and visual-inertial sensor fusion, then position attitude estimation can be achieved, but image blur correction fails when the parallax amount between images is small
Solution Approach 1:
The patent changes the selection criterion from requiring large parallax amount to selecting feature points based on their contribution to three-dimensional coordinate estimation accuracy. This parameter change allows the system to work effectively even when parallax amount is small, by focusing on feature points that provide meaningful depth information through the depth map estimation process
Solution Approach 2:
The patent replaces the traditional SFM-based mechanical tracking system with a depth map-based estimation system. Instead of relying on parallax-driven feature point matching, the system uses depth information from the depth map to directly estimate three-dimensional coordinates, substituting the mechanical parallax-based approach with a depth-information-based approach that works regardless of parallax amount
2Productivity
If images are selected to increase parallax amount, then three-dimensional coordinate estimation can be performed, but processing periods occur where no correction is performed due to small parallax
Solution Approach 1:
The patent ensures continuous image blur correction by always selecting feature points that contribute to depth map estimation, regardless of parallax amount. The system continuously performs three-dimensional coordinate estimation using the depth map approach, eliminating gaps in processing where no correction would otherwise occur
Solution Approach 2:
The patent introduces the depth map as an intermediary element that enables three-dimensional coordinate estimation without relying on parallax amount. The depth map serves as a mediator that provides depth information independently of the parallax between images, allowing continuous and precise coordinate estimation
Data Source
AI summary
An image blur correction device tracks a feature point in a captured image, calculates feature point tracking information, and estimates a positional relationship including a depth between a subject and an imaging apparatus as three-dimensional coordinates of the feature point on the basis of the feature point tracking information and position attitude information of the imaging apparatus calculated on the basis of shake information relating to shake of the imaging apparatus. The image blur correction device also controls a degree of contribution of the feature point to the estimation of the three-dimensional coordinates. The image blur correction device corrects the position attitude information according to a result of the estimation of the three-dimensional coordinates of the feature point, calculates a target position of a shake correction lens based on the shake information and the position attitude information, and drives the shake correction lens.