Progressive Image Denoising using Fast Noise Variance Estimation

The patch-less Progressive Image Denoising(PID) is physical process of reducing the noise in image based on deterministic annealing i.e. temperature decreases from high to low so that shape of kernel changes according to it. The results of PID implementation are good and excellent for both natural and computer generated images i.e. artificial or synthetic images. It estimate the noise using robust noise estimation. PID algorithm is only for denoising additive white Gaussian noise(awgn). For using PID the requirement is original image (noise free image) and amount of noise added to it. In real scenario, it is not possible to get the knowledge of noise level available in any image. This paper gives an approach to automatically estimate the noise level in the given input image and then denoise the image using PID. Experimental results demonstrate that proposed algorithm outperforms both objective and subjective fidelity criteria in image denoising.