Fuzzy rough subset method with region based mining to improve the retrieval and ranking of real time images over larger image database

摘要

Region based image mining is considered as an interesting approach that divides the images into several regions, where the features are extracted out from it and the set of features represents the contents of image from database. However, feature dimensionality and space complexity is one of the big issues in Image Retrieval Based on Content (CBIR). In this paper, fuzzy neighborhood rough subset method is used for feature reduction in an image. This helps to reduce the irrelevant features related to given query. The Support Vector Machine (SVM) is further used with fuzzy rough subset method to classify the images related to given query. This extracts well the spectral data characteristics between the query and database images. Performance of proposed fuzzy rough subset method with SVM classifier is tested against conventional methods. The results proves that the proposed method attains better classification of hyper spectral images than the other methods.