STATISTICAL ANALYSIS OF FREQUENCY DOMAIN FILTERS FOR IMAGE DENOISING

With the advent of the digital world, the use of digital images has become widespread. During the process of image acquisition, image contamination by noise becomes an inevitable part of the image, leading to a significant reduction in quality. Traditionally, filters remove noise from images. Technological advances in the arena of image processing have led to growth of more efficient techniques of image denoising. These techniques make use of wavelets and Curvelet transforms, which can be blended with known parts of noisy image to estimate the unknown parts of the image better. In this paper, we apply various image denoising techniques to a random image that is corrupted by either Gaussian noise, speckle noise or salt and pepper noise. The evaluation results in terms of Mean Squared Error, Peak Signal to Noise Ratio, Structural Similarity Index and Computation time will help to decide the best technique that is used for denoising an image distorted by noise.