Application of the polarimetric matched image filter to the enhancement of characteristic features of POL-SAR imagery

The authors investigate the potential of an exclusively polarimetric image filtering method which takes full advantage of matrix data provided on a pixel-by-pixel basis, and complements the existing scalar contrast optimization and speckle reduction techniques. The goal is contrast optimization without the help of incoherent averaging over pixels/look. In order to improve the efficiency of the polarimetric filtering method, one must combine it with other image processing and statistical communication theory techniques. In particular, the filter efficiency depends critically on peak sharpness of the histograms and one may use multilook incoherent averaging, block discretization, and nonuniform quantization to improve ´peakedness.´ Also, a continuous mapping in three-color space of the three independent polarizations (e.g., HH, VV, VH) can be used in combination with the polarimetric matched filter to enhance visual detectability of the polarimetric rough surface scattering dependence without destroying the geometric integrity of the image. Finally, an incoherent polarimetric imaging via the 4*4 real Mueller matrix may lead to an even more efficient speckle reduction procedure, in particular, for multilook and multifrequency images.<>