Image segmentation from multi-look passive polarimetric imagery

A passive imaging polarimeter records the polarization state of light reflected by an object that is illuminated with an unpolarized and usually uncontrolled source. Passive polarimetric imagery has shown to be useful in many remote sensing applications including shape extraction, material classification and target detection/recognition. In this paper, we present an image segmentation algorithm that automatically extracts an object from multi-look passive polarimetric imagery. The term multi-look refers to multiple polarization measurements where the position of the source of illumination (typically the Sun in passive systems) changes between measurements. The proposed method relies on our previous work on estimating the complex index of refraction and reflection angle from multi-look passive polarimetric imagery. We experimentally showed that the estimates for the index of refraction were largely invariant to both the position of the source and the view angle. Consequently, we utilize the index of refraction as a feature vector to design an illumination invariant image segmentation algorithm. A clustering approach based on the classic c-means algorithm is used for segmenting objects based on their index of refraction. The proposed segmentation approach is validated by using data collected under laboratory conditions. Experimental results indicate that the proposed method is effective for segmenting various targets of interest.