Moving Object Detection in Noisy Images

Abstract

The new method for detecting moving objects in an image distorted by white Gaussian noise is proposed. The method is based on the representation of the video sequence by a three-dimensional discrete Markov process and includes two stages. The first step is a three-dimensional nonlinear filtering of images which allows keeping the contours of objects. The second stage comprises detecting the objects of interest based on the assessment of their boundaries and brightness. The method requires small timing and computational expenditures which makes it possible to use this method in real-time data processing.