International Research Journal of Engineering and Technology (IRJET) Volume: 04 Issue: 04 | Apr -2017
www.irjet.net
e-ISSN: 2395 -0056 p-ISSN: 2395-0072
Image Based Information Retrieval Sneha A. Taksande1, Prof. A. V. Deorankar2 PG Scholar, Department of Computer Science and Engineering, Government College of Engineering, Amravati, Maharashtra, India.1 Associate Professor, Department of Information Technology, Government College of Engineering Amravati, Maharashtra, India.2 --------------------------------------------------------------------------------------------------------------------------------------------------in both academia and industry in recent years. Besides the Abstract-In today’s time retrieving the relevant information from huge collection of data has attracted a lot of attention. Various search system are available for that purpose but they should be able to find the most relevant search results according to users query that satisfies the users need. Different techniques are also there to retrieve these information. In traditional search engines usually text documents are considered and the images in these documents are ignored. Generally pictures in the HTML web pages are used to retrieve the other relevant pictures by comparing its textual as well as visual contents. Also in traditional text-based search engine relevant images can be retrieve using visual features by giving a textual query. Various systems and search engines are available for easy access and retrieval of relevant multimedia content. Most of them rely on textual data associated with the visual contents. So this paper present an approach that will produce the search results by considering the details of the images in web pages. Keywords: Search System, Web pages, Relevant Search Results, Images.
1. Introduction As there is the tremendous growth of internet over the past few years, a large repository of data covering almost every area has been formed over the web and as a result of which search engine users are facing a lot of problems in retrieving the most appropriate information out of it, which is known as information overkill problem. So the search system designed should be able to retrieve the good and necessary information that will fulfills the user’s needs. Due to the success of information retrieval, most commercial search engines employ text-based search techniques for image search by using associated textual information, such as file name, surrounding text, URL, etc. Even though text-based search techniques have achieved great success in document retrieval, text information is often noisy and even unavailable. In order to improve search performance, image search re-ranking, which adjusts the initial ranking orders by mining visual content or leveraging some auxiliary knowledge, is proposed, and has been the focus of attention
© 2017, IRJET
|
Impact Factor value: 5.181
|
well-known semantic gap, intent gap, which is the gap between the representation of users’ query/demand and the real intent of the users, is becoming a major problem restricting the development of image retrieval in image search re-ranking. Most of the existing re-ranking methods utilize the visual information in an unsupervised and passive manner to overcome the “semantic gap” i.e. the gap between the low-level features and high-level semantics. However, when measuring image similarity and typicality, conventional re-ranking approaches only consider visual information and initial ranks of images, while overlooking the influence of click-through data. Though multiple visual modalities can be used to further mine useful visual information that can only achieve limited performance improvements. This is because these re-ranking approaches neglect the “intent gap”. Users’ real search intent is hard to measure and capture without users’ participation and feedback. Therefore some researchers attempt to integrate users’ interaction with the search process. There are a widely accepted assumption and a generally applied strategy for most image search re-ranking approaches respectively, i.e., visually similar images should be close in a ranking list, and images with higher relevance should be ranked higher than others. In this paper we propose a novel document retrieval approach that uses the content of the pictures in the Web pages to boost the accuracy of search engines. This paper gives an approach for information retrieval based on the images. The text details of these images will be taken into account in order to compare them with the query keywords and provide the search results. Consequently our hope is that a search system will considers the textual information extracted from the pictures will yield improved accuracy of search system.
2. Related work In this paper [1] it addressed the issue of leveraging clickthrough data to reduce the intent gap of image search. They propose a novel image search re-ranking approach, named
ISO 9001:2008 Certified Journal
| Page 1393