Distributed Algorithms for Image Classification and Content-Based Retrieval using Image Classes

The recent phenomenal increases in the amounts of image data being generated and consumed in a variety of applications have necessitated the development of image databases. The richness of content, inexact nature, and subjective interpretations of images have rendered keyword-based queries inadequate. Thus image databases need to provide content-based queries and retrievals. Fast and accurate retrievals for content-based queries depend on e cient and e ective indexing and searching schemes. The classi cation of images into di erent categories (classes) is an important task which facilitates the indexing and searching process. Support for content-based queries and retrievals place enormous demands on computation due to the huge data sizes of images and the enormous number of images in typical image databases, coupled with large and complex indices required for search. Network of workstations (NOWs) are a cost-e ective way of providing the much needed computational power in such applications. This paper presents a distributed scheme for the classi cation of images in an image database, using a network of workstations. It then presents a distributed scheme for image retrievals using a NOW system. The scheme uses an initial classi cation of images in the database. It uses a heuristic for determining the centroids and thresholds for the image classes, which are used in the search process. The results of the classi cation and retrieval, and the speedups obtained using a network of workstations are presented. The results indicate the viability and e ectiveness of the proposed schemes.