Image Separation Technique and Application Based on Independent Component Analysis

吴小培,冯焕清,周荷琴,王涛

DOI: https://doi.org/10.3969/j.issn.1006-8961.2001.02.006

2001-01-01

Abstract:Independent Component Analysis(ICA) is a novel approach of blind source separation and has received attention because of its potential application in signal processing such as in speech recognition system, image processing, telecommunication and medical signal processing. In this paper, the fundamental theory and algorithm of ICA is introduced. The feasibility of ICA for sequential image processing is studied and a novel ICA based approach of moving target detection is proposed. One of the main tasks of moving target detection is how to remove the interference of background efficiently. In our research work, we find that ICA has the good performance for the background rejection. The steps of our method are that: firstly, we apply ICA to the sequential images of containing moving target to get the mixing matrix of ICA linear mixing model and the vector of the independent components. The result of ICA show that the background component is separated from other independent components. So, secondly we can set the independent components that contain obvious background image to be zero and then do the inverse computation of ICA using the mixing matrix. After the two processing steps above, we get the clear track of moving target. The experiment results illustrate that ICA has good performance for blind source separation, and the novel approach of moving target detection based on ICA is more effective than the traditional detection method.