Evaluation Web Image Search Engines (ISE)
Salam R. Al-E’mari
Department of Computer Sciences Yarmouk University
Supervisor: Belal Mustafa Abuata
Today search engines have become the most important way to information retrieval through the World Wide Web. Information has expanded greatly may consist of text, file, web page, image and other type. Images one important species in information retrieval, many users care about image retrieval from search engines where web image retrieval is a challenging task that requires efforts from image processing, link structure analysis, and web text retrieval. This paper focuses current technologies in web image search engines and compares result between the main search engines (namely Google, Bing and yahoo) depend on stander evaluate precision and recall where Google image get the highest precision through evaluate.
Key-words image retrieval, evaluation, CBIR, text-based.
User use web to get any type information all over the world, maybe information with different format URL, text, document word, PowerPoint, image, video …etc. Today there is more than one web search engines help user to information retrieval such as Google search engine, but not all information retrieved through search engines is relevant, it need filter by user. Photo one of the most species of interest to the user to process information retrieval. Modern, there are search engines especially for the process of image retrieval from large databases called Image Search Engine. Image retrieval system concern searching and retrieved digital image from collection databases. Branches of computer science interested in the process of image retrieval are databases management and computer vision. More strategies to process image retrieval, current image retrieval system use two main categories text-based image retrieval and image content-based. Text-based image describe image content by text, content-based image retrieval techniques used visual features to describe content of images (color, shape and Texture) use to automatic image annotation . Web image search engine is tool cans any user use to search on image. ISE indexing categorizes in three ways, first way text-based index through caption, file name, surround text, and text in the HTML. Second way image-based index through visual features image color, texture, and shape. Third way is hybrid of text and image index. Web image search engine provide box (input text interface) that user can write query by
keywords. Then query is process by matched against the indexed web images then return list of relevant result order by rank descending. Image understanding is still an open research problem . A fundamental problem with scientific uses search engines is that their results can change overtime, more documents add and remover through time. This problem the same for web image search engine so the compared observations tasked based on three rules. First compared observations are performed at the same time, or repeat as regular interval which leads respectively to instantaneous fluctuations and to fluctuations detected through time. Second compared observations consist of just one query, or of a set of several queries which leads respectively to fluctuations from singular observations and to fluctuations from plural observations. Third compared observations are the same query or they are different which leads respectively to fluctuations from the same observation and to fluctuations from different observations . This paper evaluate the extent to which search engine results in web image investigations three engines Google, Yahoo and Bing dependent through a comparison of the results across a set of queries and use stander evaluate precision and recall. Next section II literature review discusses current mechanisms in web image search engines. Section III Goal and approach evaluate...
Please join StudyMode to read the full document