Method of estimating blur kernel from edge profiles in a blurry image
US8594447B2 · kind B2 · utility
Assignee
Inventor
Key dates
| Filing date | Dec 23, 2011 |
| Grant date | Nov 26, 2013 |
| Priority date | — |
| Expiry date | Jul 20, 2032 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06T2207/20076
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
A method of deblurring a two-dimensional, blurred image. An edge within the blurred image is detected, and an edge profile of the blurred image is extracted. A length of the edge profile is determined, and a blur kernel size is estimated based thereon. A quantile function is applied to the edge profile, thereby generating quantile function output. A linearity test is performed on the quantile function output, thereby obtaining a preliminary linearity metric. Another quantile function is selected and the applying a quantile function to the edge profile, thereby generating quantile function output, and the performing a linearity test on the quantile function output, thereby obtaining a preliminary linearity metric, is iteratively repeated until a best linearity metric, a best quantile function, and a best quantile function output are determined. A slope of the best quantile function output is determined, and a blurring parameter is determined based thereon. A blur kernel is estimated based on the blurring parameter and the blur kernel size. A sharp image is resolved using the blur kernel.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.