Image Denoising Using Weighted Averaging

Guleryuz proposed a simple and powerful image denoising algorithm using weighted averaging based on DCTs. The shortcomings of Guleryuz´s method are that it needs to train two threshold parameters and its denoising ability deteriorates when noise level becomes high. In this paper, we give a method which trains the two parameters. We also improve Guleryuz´s method via local Wiener filtering. Our method only needs to train a threshold parameter and also performs significantly better than Guleryuz´s method.