Motion Blur Direction Estimation via Iterative Test Blurring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for reducing motion blur in images often require knowledge of the point spread function (PSF) or a good initial guess for blur parameters, which can be challenging to obtain, especially when estimating the blur direction.
Innovation Solution
A method that blurs an image in multiple test directions to determine the direction resulting in the smallest change, identifying the test direction that is most similar to the actual blur direction by comparing differences in image appearance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If deconvolution methods are used to reduce blur in a blurry image, then image quality is improved, but the method requires the point spread function (PSF) to be known or automatically estimated
Solution Approach 1:
The system uses the blurred image itself to estimate the PSF through iterative deconvolution, rather than requiring external calibration data or manual PSF input. The blurred image serves as both the input and the source of information for determining the blur characteristics.
Solution Approach 2:
The method implements an iterative process where the estimated PSF is used to deconvolve the image, the result is evaluated, and the PSF estimation is refined based on the evaluation feedback. This closed-loop approach continuously improves the PSF estimate until convergence.
2Manufacturing precision
If PSF estimation methods are used, then blur reduction is enabled, but a good initial guess for blur parameters such as blur direction is required
Solution Approach 1:
The system performs preliminary analysis of the blurred image to generate an initial guess for the PSF and blur parameters before initiating the main deconvolution process. This preliminary step provides a starting point that is already reasonably close to the true values, improving convergence.
Solution Approach 2:
The blur parameter estimation is made dynamic and adaptive throughout the iterative process. The initial guess is refined in successive iterations based on the actual image content and blur characteristics observed, allowing the system to adapt to different types and degrees of motion blur.
3Measurement precision
If multiple test directions are evaluated to estimate blur direction, then estimation accuracy is improved, but computational complexity increases
Solution Approach 1:
The evaluation of test directions is segmented into a discrete set of candidate directions rather than evaluating all possible directions continuously. This segmentation reduces the search space while maintaining sufficient accuracy for practical applications.
Solution Approach 2:
The method evaluates a sufficient number of test directions to achieve accurate blur direction estimation without exhaustively testing every possible direction. This partial action approach achieves the necessary precision with reduced computational burden compared to complete enumeration.
Data Source
AI summary
A method for estimating a blur direction (20) of motion blur (16) in a blurred image (14) includes the steps of blurring the blurred image (14) in a number of different test directions (360A) (362A) (364A), and finding the test direction (360A) (362A) (364A) for which the blurred image (14) changes the least by the additional blurring (366). With this design, when more blur (366) is applied to the blurred image (14) in a test direction (360A) (362A) (364A) that is similar to the blur direction (20), the difference in the image appearance is relatively small. However, when more blur (366) is applied to the blurred image (14) in a test direction (360A) (362A) (364A) that is very different to the blur direction (20), the difference in the image appearance is relatively large. In one embodiment, a blur difference is determined for each test direction (360A) (362A) (364A). Subsequently, the test direction (360A) (362A) (364A) with the smallest blur difference is selected as the blur direction (20). Alternatively, the estimated blur direction (20) can be perpendicular to the test direction (360A) (362A) (364A) with the largest blur difference.