A Comparative Study of Distance Measurement Performances for Image Retrieval using Clustering Techniques

Abstract: Image retrieval is the dealing out of looking and retrieving photographs from a huge dataset. As the images grow complex and varied, retrieval the right images becomes a problematic challenge. For centuries, many of the images retrieval is text-based which means searching is based on those keyword and text generated by human‟s creation. On the whole a content based image retrieval system will retrieve some of the aspects image like shape, texture, color and spatial information of each image which is placed in the database and then stores the feature details in a different database called the feature database. The characteristic database carries the function facts of all the snap shots present inside the major database. The feature records are very small in length whilst in comparison with the original picture. The feature database holds the description of the main image in a Compact format. It holds information about the coloration, shape, texture and spatial statistics in a set period actual-valued multi thing function vectors or signature. We can retrieve the feature vectors based on similarity measurements. In this paper, we can survey various similarity measurements in K-means clustering to retrieve the images from image database and a comparison study is also made on the various measurements using precision, recall and F1 score.

Keywords: Content based image retrieval, feature vectors, similarity measurements, and clustering approach.