-
Method. For all blocks ![](http://www.cs.tut.fi/~dabov/3D-DFT/swptransparent.gif) from the noisy image, we do the following steps.
-
Block-matching. The result of block-matching is a set ![](http://www.cs.tut.fi/~dabov/3D-DFT/swptransparent.gif) of the coordinates of the blocks that are similar to the currently-processed reference block, ![](http://www.cs.tut.fi/~dabov/3D-DFT/swptransparent.gif) , according to our ![](http://www.cs.tut.fi/~dabov/3D-DFT/swptransparent.gif) -distance measure;
![MATH](article_v6__19.png)
-
Denoising by hard-thresholding in local 3D transform domain.![MATH](article_v6__20.png)
-
![](http://www.cs.tut.fi/~dabov/3D-DFT/swptransparent.gif) comprises of ![](http://www.cs.tut.fi/~dabov/3D-DFT/swptransparent.gif) stacked local estimates ![](http://www.cs.tut.fi/~dabov/3D-DFT/swptransparent.gif) of the true image blocks located at the matched locations ![](http://www.cs.tut.fi/~dabov/3D-DFT/swptransparent.gif) .
-
![](http://www.cs.tut.fi/~dabov/3D-DFT/swptransparent.gif) is a weight which is inversely proportional to the number of non-zero transform coefficients after hard-thresholding, ![](http://www.cs.tut.fi/~dabov/3D-DFT/swptransparent.gif) .
|