Image Classification using Frequent Itemset Mining

Abstract

Image classification is one of the most useful and essential research field in computer vision domain and challenging task in the image management and retrieval system. The growing demands for image classification in computer vision having application such as video surveillance, image and video retrieval, web content analysis, biometrics etc. have pushed application developers to search and classify images more efficiently. The main goal of image classification is to classifying image into different classes according to their visual characteristics. In this paper, we propose an approach for image classification by applying frequent itemset mining. Frequent Itemset Mining is used to finding the frequent patterns which is referring as Frequent Local Histograms or FLHs. All Local Histogram information must be maintained for obtaining these FLHs patterns and demonstrate the Bagof-FLHs based image representation. The main aim of this research is to use PCA-SIFT (Principal Component AnalysisScale Invariant Feature Transform) local descriptor for feature extraction. It is more distinctive, robust to image deformations and more compact than the standard SIFT descriptor. The proposed work reduces the overall time of image classification task and maintaining its overall accuracy. General Terms Image Classification, Frequent Itemset Mining (FIM)