VISUAL AND TEXTUAL FEATURES BASED IMAGE SEARCH RESULTS AND PREFERENCE LEARNING MODEL FOR QUALITY ASSESSMENT

Abstract

Retrieval of images from database plays a significant role in day to day life. Widespread investigation methods have been studied in recent work for retrieving
images which is relevant to a user given query. Several factors are able to control of image search results. Concerning different settings in favor of these factors produce search result lists by means of varying levels of quality. On the other hand, no setting is able to constantly execute optimal results for all user given queries. So, known a set of search result lists produced through different settings, it is important to repeatedly decide which result list is the optimal in order to
present it to users. At the same time to improve the efficiency of the system the major steps of image retrieval system such  as feature extraction, similarity
measurements and retrieval or matching is also discussed in this paper. Those steps extraction of the texture features and visual features increase the efficiency of retrieval system, so in this research work texture feature extraction is done using Grey Level Co-occurrence Matrix (GLCM). The major contribution of the paper consist of three major steps: at the initial stage of the work, a classification model based on user preference is formulated to find the most excellent image search result list classification problem. At the second step most important features such as visual and texture features are extracted from the  returned images. Third, a query based preference classification model is created based on the user specified query. The results of third step have been tested on a diversity of applications such as the reranking capability evaluation, optimal search engine collection, and synonymous query proposition