Improved denoising technique for natural and synthetic images

Image processings applications like in object tracking, medical imaging, satellite imaging, face recognition and segmentation requires image denoising as the preprocessing step. Problem with current image denoising methods are bluring and artifacts introduces after removal of noise from image. Current denoising methods are based on patches of image has well denoisinging ability but implemention of such methods are difficult. The patch-less Progressive Image Denoising(PID) is a dual-domain image denoising method which progressively removes reduce the noise from image in each iteration. It has simple implementation using robust noise estimation and deterministic annealing. Its results are artifacts free. It is better for the artifical images i.e. computer generated images or synthetic images. This paper presents comparatively results of PID, Dual-Domain Image Denoising (DDID) and Block Matching and 3D Filtering (BM3D) for both natural and synthetic images contaminated with different levels of AWGN noise.