Improving Denoising Performance with Quality Enhancement Using Spatially Adaptive Iterative Filtering

Abstract

Digital image processing is a challenging domain of programming. There is always some noise present in all digital images. When we have to remove zero-mean white and homogeneous Gaussian additive noise from a given image, we address it as Image denoising problem. This is an important pre-processing task, for which spatial domain and transform domain image filters have achieved great success. However we cannot fine tune the denoising strength using spatial domain filters, but it can be efficiently done using shrinkage operators (in transform domain). In this work, we are proposing a novel approach for controlling the denoising strength using Spatially Adaptive Iterative Filtering (SAIF). The highlight of this technique is that we can automatically optimize the type of iteration and the iteration number w.r.t estimated risk using the plug-in risk estimator, after the adaptive iteration of filtering local image content with given base filter. Improved performance than often employed SURE estimator is the attracting characteristic of plug in estimator. Finally it is extended with guided filtering for quality improvement. Experimental results prove that SAIF with guided filtering improves denoising performance undoubtedly.