Decomposition-based shape template matching for CBIR system

Abstract

Content Based Image Retrieval is a process to get a desired image from a substantial database. We propose a template for shape based hierarchical feature matching approach for content based image retrieval system. It utilizes a combination of global feature for shape based templates. In this work a new learning method is put forth which is based on the hierarchal decomposition of the data. The proposed method establishes learning algorithm where the feature extraction process is executed to detect edge, orientations and shape of the dataset images. Thus extracted shape based features are used for matching the template to improve the retrieval accuracy. The proposed model is tested for the Wang dataset. The classification of dataset is taken care by the support vector machine algorithm with the accuracy of 99.09 % The retrieval results of proposed model is illustrated in terms of precision and recall, the improved efficiency of retrieval is compared to other existing models.