Comprehensive Analysis of Various Distance Metrics on Colour-Based CBIR System

Abstract

AbstractIn the recent decades, lots of images have been added to the database and is growing rapidly. Since the database is very huge, hence retrieval and querying of these images become difficult. The content-based image retrieval (CBIR) system provides an efficient option, that is based on extraction of image features and compare. The primary method of feature extraction in CBIR system is based on the colour content of images. Finding the similarity among image features is mainly based on distance metrics, and it plays crucial role in image retrieval. There are many such similarity metrics found from the literature, few of them perform well on some specific cases. Thus, it is important to know, the appropriate metrics for optimal image retrieval. This article presents a comprehensive survey on various popular distance metrics on wide range of image database. The survey is based on finding the similarity of image features, that are based on colour content of an image. The survey gives good insight into the similarity measuring metrices. The popular set of large image database is used for analysis purpose. The results show, superiority of Canberra and Bray–Curtis distances over other distance metrics.