文章基本信息

标题：CLUSTERING SIMILAR IMAGES USING VARIOUS IMAGE FEATURES
本地全文：下载
作者：Dr. E. S. Samundeeswari ; M. Kiruthika
期刊名称：International Journal of Advanced Research in Computer Engineering & Technology (IJARCET)
印刷版ISSN：2278-1323
出版年度：2018
卷号：7
期号：3
页码：251-257
出版社：Shri Pannalal Research Institute of Technolgy
摘要：Clustering is an unsupervised classification of patterns into groups (clusters). The images can be clustered into group of visually similar images and can be used in applications like extracting images similar to the query image. The proposed system uses descriptors such as pixel information, foreground/background features, texture features for grouping similar objects. The cluster validation methods applied are internal and stability measures. It includes Connectivity, Silhouette, Dunn index in internal measures and Average Proportion of Non-overlap (APN), Average Distance (AD), Average Distance between Means (ADM), Figure of Merit (FOM) in Stability measures. These methods are used to compare multiple clustering algorithms such as K-Means Clustering, Hierarchical Clustering and PAM clustering for identifying the best clustering approach and the optimal number of clusters. Out of the three Clustering methods, K-Means Clustering provides the best cluster result. The texture feature produces better cluster result when compared to pixel information, foreground/background feature and combination of both texture and foreground/background feature. The system is user friendly and developed using „R‟.
关键词：Clustering; Descriptors; Hierarchical; K- Means; PAM; Silhouette Cluster Validation; Textures.