A resynthesis framework for PolSAR images based on feature selection

Compared with optical remote sensing images, PolSAR images is hard to be interpreted by human eyes. In order to improve the visual interpretability and automatic interpretation, an effective resynthesis framework is developed in this paper to highlight the interesting targets for PolSAR images. The resynthesized images are constructed by the method similar with polarization synthesis which inputs different combinations of polarization parameters extracted from original PolSAR images into three channels of a display space. Since almost complete polarization information extracted from PolSAR images has been used in our framework, feature selection method is needed to choose the most essential parameters for resynthesizing. Experimental results demonstrate that our resynthesized PolSAR images are more readable and more intelligible.