Research On Dominant Descriptor Selection Based Mobile Visual Search Algorithm

With the rapid increase amount of smart phones and other mobile devices, people have the intention to use these mobile devices to take pictures and search them on the internet anywhere. As a kind of interesting mobile application, Mobile Visual Search (MVS) has attracted extensive research efforts from both academy and industry.Most of the MVS systems adopt the client-server framework, in which transmission latency caused by the limited bandwidth in wireless network is a big problem. To address this problem, the state-of-the-art work focuses on designing low bit-rate descriptors for MVS. However, little work focuses on reducing the number of descriptors. To further reduce the latency, we propose a novel framework of MVS based on the weighted matching of dominant descriptor.Firstly, we present an affinity propagation based algorithm for dominant descriptor selection. Besides, we propose a weighted feature matching method to consider the differences of dominant descriptors in feature matching. By the proposed framework, we not only reduce the network latency in MVS, but also avoid transmitting useless descriptors to improve the retrieval accuracy of MVS.Secondly, there exit some complicated image containing large amount of descriptors, which require massive computing ability and storage space, which are both limited on mobile devices. In such situation, we further proposed a sparse representation based model to address this issue. We first sampling some of the given descriptors to select dominant descriptors, then for the out of sample group descriptors, we adapt sparse representation based method to classify them. By such processing, we manage to reduce the computing and storage cost. Then we can continue to do the remaining work.The experiment results on Stanford MVS data set show that when using CHoG descriptors, the proposed framework outperforms the existing framework by reducing more than 40%of the amount of data transmission and increasing 5%of the average retrieval accuracy. Experiment results on Oxford Building data set show that when facing complicated images, the improved algorithm can effectively select dominant descriptors and acquire competitive retrieval accuracy with much less computing and storage cost.