A Hybrid Filtering Algorithm Based On Advanced D-S Theory Of Evidence

Image signal is often interfered by many kinds of noise when produced, transmitted and recorded. The quality degradation of image must be preprocessed before edge detection, image segmentation, feature extraction, pattern recognition.In the preprocessing, image filter is very important. Linear and nonlinear filters are in common used. Linear filter can filter gauss noise, but pulse noise and high frequency can not be filtered. On the contrary, nonlinear filter can filter pulse noise, but can not filter even distribution noise and gauss noise. So we can combine them in practice.Information fusion technology is a method of information processing and analyzing. It can process the integrated information. This method is more credibility than simple information processing, so it is a regarded research way. It will attract people's attention to combine information fusion and image processing technology.The D-S theory of evidence belongs to decision-level fusion in information fusion. It combines some evidence to make decision, so this is a kind of decision-level fusion. It not only can treat with the uncertainty brought by the uncertainty of information, but also can treat with the uncertainty brought by the ignorance. It can distinguish the difference of the uncertainty and the ignorance.This dissertation firstly discusses the image filter, and analyses both mean filter in linear filter and median filter in nonlinear filter. Then they are combined in the hybrid filter.In the hybrid filter, there are some disadvantages of great risk of...