Wavelet based iterative thresholding for denoising of remotely sensed optical and synthetic aperture radar images

This article presents an overview of various denoising methods for optical and synthetic aperture radar (SAR) images. Currently, there are numerous algorithms and techniques to denoise images using adaptive filters in both the spatial as well as wavelet domain. However, in contrast, an algorithm which performs both soft and hard thresholding on a multi-level wavelet transformed image utilizing an adaptive threshold value has been designed and implemented. The threshold value varies for different wavelet regions from image to image and is selected for each region based on certain performance criteria such as Peak Signal to Noise Ratio (PSNR) and Mean Square Error (MSE). This algorithm has been used to denoise SAR images corrupted with multiplicative noise as well as optical images corrupted with White Gaussian additive noise. The results obtained have been compared with existing filters like median filter, Frost filter and Wiener filter. Additionally, we have compared the results of using different wavelet families including family of Daubechies and Biorthogonal filter banks.