An efficient approach for monuments image retrieval using multi-visual descriptors

Abstract

This paper presents retrieval of monument images using low-level descriptors built on color, texture, and shape retrievals using CBIR technology with the help of MATLAB tool. Representation of shape is very complex because of this shape gains special importance among all these features. Morphological operators are used for the extraction of shape-based descriptors, improved local binary pattern (ILBP) is used for the extraction of texture-based descriptors, and RGB color histogram is used for extracting color-based descriptors. Morphological gradients are obtained using morphological operators, and then moment invariant is applied on these gradients. ILBP is used because it discovers the group of elementary primitives like lines, cross-intersections, and T-junctions that are unnoticed by uniform LBP method. ILBP feature used here is more precise than traditional LBP descriptor. Stout set is built to find and retrieve the images of similar kind. Tests are performed on the database having 360 images with six categories. Results show that the proposed system is capable of retrieving alike images more accurately.