Indian Journal of Science and Technology

Abstract

Objectives: The main objective of this paper is to retrieve Near Duplicate images from set of images similar to that of the user’s query image. Methods/Statistical analysis: Initially, the features are extracted from all the images in the image set and clustering is done using K-means and PSO. Then the similarity is calculated between the query image features and the features of images in the set. California ND image dataset is used for testing purpose. Initially, around 20 images are taken and tested. Findings: In this paper, clustering of near duplicate images is done along with optimization technique i.e., PSO is used with k means clustering. The experimental study indicates that the performance of the proposed system is much better when compared with other present systems that are not using optimisation methods. Application/Improvements: Near Duplicate Detection has received more attention in the recent years due to applications in copyright enforcement, organizing large image databases, easy image search, elimination of duplicate logos, etc. K means - PSO Clustering provides faster search-by-query images, image grouping and facilitates users’ browsing.
Keywords: Clustering, Image Retrieval, K Means, Near Duplicates, PSO