Robust nonlinear filtration of noisy-image sequences

Abstract

1. Nonlinear filtration methods are extensively used in remote sounding, processing of medical information such as electroencephalograms, cardiograms, or tomograms. These methods suppress the effect of noise of various physical natures [1‐3]. In this work, we proposed a new approach based on the joint use of rank statistics and M-estimates (generalized maximum-likelihood estimates) [4]. This approach enables one to use the advantages of each method of nonlinear filtration and to develop a new class of robust algorithms of filtering both individual images and sequences of exposures (video and tomographic). This provides a significant improvement in the quality of filtering images and sequences of video exposures, including the suppression of various types of noise, reconstruction of small details of images, and on-line realization of filtration procedures. The latter is particularly important and realized through algorithms based on the digital processing of signals with the use of a TMS320C6701 digital signal processor [5]. This experimentally corroborates the possibility of on-line filtering of various sequences of video images in the presence of various types of noise.