Image to Image Search using K-means Clustering

Abstract

With the increase in the amount of information on the World Wide Web, it has become a difficult task of finding relevant information. More often, textual based search engines were used but there were very few search engines using which similar images can be searched all over the internet with an image as input. This method of finding information is called Content-Based Image Retrieval (CBIR). Commonly used engines for CBIR like Google Images, Yahoo! Images are based on textual annotation of Images. The images that form the result have to be previously tagged in order to appear as result. In this research, K-means clustering for segmentation of images into clusters have been studied and applied, thereby finding similar images without the need of labelling or tagging images in the database.