CLUSTERING TECHNIQUES FOR CONTENT BASED IMAGE RETRIEVAL USING MATHEMATICAL MORPHOLOGY

Abstract

Clustering Techniques For Content Based Image Retrieval Using Mathematical Morphology" is to check the applicability of the Mathematical Morphology in Content Based Image Retrieval (CBIR) to extract the derived features of the images and to retrieve to relevant images from the image database using efficient Data Mining Clustering algorithms. The present Proposed System uses CBIR technology to retrieve the relevant images from the image database on the basis of derived feature such as color, size, shape and texture. Before CBIR was developed, the images were retrieved from large databases by appending some index or address to the images. But this did not work efficiently when user queries consisted of primitive, logical and abstract features. The purpose of this dissertation work is to check the applicability of the Mathematical Morphology in Content Based Image Retrieval to extract the shape feature from the images using Morphological operation pattern spectrum as the pattern spectrum acts as a shape descriptor for extracting features from the images. Data-Mining clustering techniques such as RObust hierarchical Clustering with linKs (ROCK) and the Clustering Using REpresentatives (CURE) were used for the image retrieval . Concept of bins was used for fast and efficient image retrieval from the image data collection.The performance evaluation of these two clustering algorithms were calculated in terms of domain size and time.