期刊名称：International Journal of Innovative Research in Computer and Communication Engineering
印刷版ISSN：2320-9798
电子版ISSN：2320-9801
出版年度：2015
卷号：3
期号：8
DOI：10.15680/IJIRCCE.2015. 0308094
出版社：S&S Publications
摘要：Images have a greatest importance in computer vision. Query image search based on visual similaritieshas been lively topic of research in recent years. Increasing use of images on the internet, there is a robust need todevelop techniques for efficient and scalable image search. Usually hashing methods are used to embed highdimensional image features into Hamming space, where search can be performed in real-time based on Hammingdistance of compact hash codes. There are large numbers of images sharing same hamming distances to a query image,so that fine-grained ranking is very important. This paper proposed query adaptive image retrieval system to retrieveimage with equal hamming distance to the query. This is achieved by firstly extracting the features of an individualimage and then features are embedded into hash codes and store in a database. Query adaptive weights are thencalculated by evaluating the proximity between a query and the semantic concept classes. In this system images arerepresented using the popular bag-of-visual words (BOW) framework, local invariant image descriptors are extractedand quantized as per a set of visual words. By using Flicker image dataset for experiments, it shows clear improvementfrom our proposed approach.
