Multiresolution denoising techniques of color images with singularities like direction, line, curve and texture

Noise in images is ubiquitous and extraction of relevant features from noisy images is non-trivial problems. For satisfactory results of any higher level image processing like segmentation, recognition, authentication, compression, etc image denoising is an inevitable pre-processing step. Most of the practical images have directional features, edges, curves, lines and texture singularities as high frequency components of the images. When denoising algorithms are implemented, it is very important that such singularities should be preserved, as these singularities are essential to retain satisfactory visual quality along with important image features. In this paper, comparative performance of some well known multiresolution geometric image transforms are evaluated for denoising color images containing singularities so as to bring out the suitability of using each transform in respect of different types of images and noises.