Blur difference estimation using multi-kernel convolution
US8564712B2 · kind B2 · utility
Assignee
Inventor
Key dates
| Filing date | Mar 15, 2011 |
| Grant date | Oct 22, 2013 |
| Priority date | — |
| Expiry date | Jan 30, 2032 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06T2207/10148
- WIPO fieldAudio-visual technology
- WIPO sectorElectrical engineering
Abstract
An apparatus and method for rapidly and accurately determining blur differences between captured images. Blur change is modeled as a point spread function from a first position to a second position, which is approximated in response to performing a series of convolutions using at least two different blur kernels having different variance. The kernel with a larger variance is used first to speed processing, after which a kernel having a smaller variance is utilized to attain desired accuracy. Any number of kernels can be utilized with decreasing variance to achieve any desired level of accuracy. The apparatus and method can be utilized within a wide range of image capture and/or processing devices, and can be utilized within camera focus mechanisms to increase focusing speed and accuracy.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.