Content based Image Retrieval using Median Binary Pattern with Structure Element Descriptor on Noisy Images

Abstract

Effective features are essential issue for the representation and recognition of images. In this paper, we present a new feature representation method for image retrieval based on Median Binary Pattern (MBP) and Structure Element Descriptor, named Median Binary Pattern extended structure element descriptor (MBP-SED). Firstly, the gray scale image is converted using the MBP. The elements in the structure can effectively represent the bitmap which is generated by MBP. Global Feature is achieved using the combination of MBP-SED and color Histogram (CH). Proposed system is more reliable on noisy images. Experiments on Corel-1k dataset show that MBP-SED outperforms than other descriptors.