A soft approach for feature selection and recognition of outdoor natural images

For applications like scene classification, CBIR, automated tagging, motion of a robot in a real world using vision system, understanding the images captured by the visual sensors is quite important. For such high level processing task it is mandatory to have a set of good feature. This paper discusses the application of fuzzy rough set theory for selecting a set of best features from the feature set of regions which are segmented out from outdoor natural images using a segmentation algorithm. The paper explains a complete system i.e. from feature selection/region labelling based on fuzzy rough set theory. The images used in experiments are color-texture natural images taken from the campus of University of Hyderabad.