Effectiveness Of Content-based Image Clustering Algorithms

Retrieval of a set of similar image documents requires clustering the images based on theirrnsimilar features. Clustered images are utilized by Content-Based Image Retrieval (CBIR) andrnquerying system that requires effective query matching in large image databases. Contentbasedrnimage clustering provides a more efficient method of management and retrieval ofrnlarge number of images documents. The Content-based image clustering facilitates users tornbrowse through only a particular subset of related image documents in an efficient manner.rnThis study focus in validating the two commonly image clustering algorithms namely:rnhierarchical and k-means. The validation is based on a set of selected MPEG-7 image featurerndescriptors. The similarity measure input to these clustering algorithms considers bothrnquantitative and predicate-based similarity measures. We computed two similarity measuresrntotal color-based similarity matrix as a weighted sum of the MPEG-7 color descriptors andrntotal similarity matrix as a weighted sum of color, texture and shape features.rnThe proposed metric to measure the effectiveness of clustering subsets of COREL colorrnphoto images is with respect to their semantic meaning. Shannon’s information theory isrnselected in the measuring the image’s cluster cohesiveness. The clusters formed are said tornbe well separated when the distinct clusters formed are associated to a specific imagernsemantic. The separation among clusters becomes better when the semantic association ofrnimages to a cluster is predictable. The intra-cluster cohesiveness is also captured by thernShannon’s entropy measure in measuring the clusters separation.rnThe best quality clusters are formed by the hierarchical method that uses the average-linkagernmethod when the same total color similarity matrix is input to all clustering algorithms.rnExperimental result shows that the quality of clusters formed by k-means clustering is notrnbetter than any of the three hierarchical methods. Hierarchical method which uses averagelinkagernproduced quality of clusters three times better as compared to k-means. Even thoughrnweighted texture and shape similarity measures were used in addition to total color thernaverage HACM is the best method compared to both the k-means in the formation of bothrn- x -rnsemantic and cluster cohesive clusters. The other different result obtained is that the additionrnof texture and shape feature degrades cluster quality for all hierarchical methods.

Get Full Work

Report copyright infringement or plagiarism