Research On Fusion-despeckle Algorithm Of Polarimetric SAR Images Using Contourlet Transform

Contoulet transform is a new dimensional representation of image, which boasts the decomposition and transform ability that shows the multi-scaled, multi-resolutional and multi directional features of geometric-information. Compared with wavelet transform, Contourlet transform has the obvious characteristics of anisotropy, so it is able to capture the singularity lying in the line and surface of the dimensional image. Its basic iadea is to first capture the edge singular points with a multiscale decomposition (Laplasse Pyramid decomposition) similar to wavelet transform, and then integrate the near singular ponits into new contour segment according to their direction. Image fusion refers to the processing procedure of two or more different images and other information of the same target got from different sensors. The key point is to process the spatially or temporally complementary or redundant data to obtain more accurate and abundant information by some certain algorithm, and to thereby obtaining information than any single data more accurate, more abundant and to get a new image of the target after comprehensive analysis.Polarimetric Synthetic Aperture Radar (SAR) is an all-weather, all-time high resolution microwave remote sensing imaging radar, which is based on the principle of coherent imaging system, and so imaging process is always with speckle noise. To deal with the coherent speckle noise, the algorithm polarimetric whitening filter (PWF) is often used beofore. The first step is to get the target polarization covariance matrix, the next is to calculate out the polarization measurement vector, and the last is to process the target SAR image according to the PWF algorithm.As the multi-resolution was raised and Contourlet transform develops, this article proposes one fusion algorithm based on the contourlet transform to SAR image fusion, so as to achieve the effect of speckle noise suppression. The algorithm goes as follows: the first step is to conduct the multi-scale and multi-directional decomposition to the target multiple polarization images by Contourlet transform and get the low-frequency coefficient sub-image and the different-resolutional and different-directional high- frequency coefficient sub-images; the next is to process the low-frequency coefficient sub-images by the average fusion method; the next is to study the contents of high frequency sub-images in the same direction, and measure regional differences of the high-frequency sub-band coefficients in every direction to get the higher and better infusion coefficients; and the last is to do Contourlet transform to the infusion coefficients; all done, the new image comes.