Wavelet Based Clutter Suppression Technique Using Hidden Markov Models

In dim point-like moving target detection and tracking problems under Strong background, because the intensity of the background clutter much larger than the target brightness, and often do not know its statistical properties , it is very difficult to achieve target tracking and the direct detection in strong background. The clutter suppression can play to get rid of background clutter furthest, retain target and noise and it is a necessary part of achieving the detector. It can transform the original image into a model of target and noise suitable to the detector and is an important manner in the position of target detection and tracking system.This article presents two perspectives of the experimental analysis of implementing image clutter suppression of the HMM based on wavelet domain. First, it is the usual algorithm of HMM based on wavelet domain, in this algorithm, the HMM model is set up by high-frequency coefficients of wavelet transform , and the optimal model parameters are obtained by the training of model parameters ,which is used to estimate the image background; in another algorithm , the HMM model is set up by the low frequency of wavelet transform to train the model parameters, which is similar to that of high-frequency coefficients approach to estimate the method of image background clutter. This algorithm of clutter suppression based on background estimation can be used to achieve clutter suppression by removing background clutter from the original image. After clutter suppression, the residual Gaussian premature and independence is validated by Kendall's rank correlation method and Friedman statistic method. Test results show the feasibility and effectiveness of this algorithm.