Harvesting Image Databases from The Web

Abstract

The research work presented here includes data mining needs and study of their algorithm for various extraction purpose. It also includes work that has been done in the field of harvesting images from web. Here the proposed method is to harvest image databases from web. We can automatically generate a large number of images for a specified object. By applying concept of data mining and the algorithm from data mining which is used for extraction of data or harvesting images. A multimodal approach employing text ,metadata and visual  features is used to gather many high-quality images from the web. The modules can be made to find query images by selecting images where nearby text is top ranked by the topic i.e., formation of image clusters then download associate images by using approaches like web search, image search and Google images. Apply re-ranking algorithm and then filtering process to harvest the images.Currently, image search gives a very low precision (only about 4%) and is not used for the harvesting experiments. Since the movements of the technologies are growing rapidly the kinds of work also need to be grown up. This work shows an approach to harvest a large number of images of a particular class automatically and to achieve this with high precision by providing training databases so that a new object model can be learned effortlessly. Many other tools also are available for harvesting images from web .An approach in this paper is original and up to the mark.

Keywords: Legacy code, re-engineering, class diagrams, Aggregation, Association, Attributes