Fusion of color histograms using PCA for SAR data classification

Color model express colors in a prescribed way, according to a certain specification. The color of image pixels could be represented in distinct color spaces which takes into consideration different properties. This paper presents the study of different color space models for land cover classification. The work is focused around generating the pseudo color image using fully polarimetric SAR data and extracting different color space histograms from that image. For taking into account the discriminating ability of all the color histograms, they are further fused using principal component analysis (PCA). The significant principal components are selected using catell´s scree test and further classified using K-means classifier. The algorithm shows an overall accuracy of around 83.5%.