Texture Based Image Retrieval Using Semivariogram and Various Distance Measures

Abstract

In a content-based image retrieval system(CBIR) feature classification,identification, and extraction play an important role. The retrieval of images using a single feature is a challenging task in CBIR systems. The high retrieval rates are reported based on combining multiple features, multiple algorithms and preprocessing steps, feature classification, and segmentation because the image retrieval are mainly based on the content in an image. This paper presents a texture feature extraction for the image retrieval system from semivariogram and robust semivariogram technique. A semivariogram is a statistical approach that provides the textural information based on the lag distance 'h'.The proposed method is tested on various standard image databases such as Corel-1k, Corel-10k, and Coil-100 database. The semivariogram and robust semivariogram methods are tested for the Corel 1k database using four distance metrics i.e. Euclidean, Manhatten, Canberra, and Chord distance to check which distance measure is appropriate for the CBIR system.The proposed method is also tested on three types of databases to investigate the performance of the CBIR system. The Matlab simulation results show that the effective performance of the system with Euclidean distance.