Method for noise reduction in digital images with locally different and directional noise
US8139835B2 · kind B2 · utility
Assignee
Inventor
Key dates
| Filing date | Mar 14, 2008 |
| Grant date | Mar 20, 2012 |
| Priority date | — |
| Expiry date | Jan 17, 2031 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06T2207/30004
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
A method is disclosed for noise reduction in images with locally different and directional noise, in particular for noise reduction in image data records of computed tomography. In at least one embodiment of the method, two image data records of an identical object region that have mutually independent noise are provided. The two image data records are decomposed by a discrete wavelet transformation into a number of frequency bands, detailed images having high frequency structures being obtained in at least two different directions. Noise images in the respective frequency bands and directions are obtained by subtracting a wavelet coefficient of the two input images. These noise images are used to estimate noise locally and as a function of direction, and on the basis of this estimate local threshold values are calculated and applied to the averaged wavelet coefficients of the input images. A result image with reduced noise is obtained after an inverse wavelet transformation.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.