PROJECT TITLE:

A Scalable Approach for Content-Based Image Retrieval in Peer-to-Peer Networks - 2016

ABSTRACT:

Peer-to-peer NetWorking offers a scalable solution for sharing multimedia knowledge across the network. With a massive amount of visual knowledge distributed among completely different nodes, it is an necessary however challenging issue to perform content-based mostly retrieval in peer-to-peer networks. Whereas most of the existing ways target indexing high dimensional visual features and have limitations of scalability, during this paper we tend to propose a scalable approach for content-based image retrieval in peer-to-peer networks by using the bag-of-visual-words model. Compared with centralized environments, the key challenge is to efficiently get a global codebook, as pictures are distributed across the whole peer-to-peer network. In addition, a peer-to-peer network typically evolves dynamically, that makes a static codebook less effective for retrieval tasks. Therefore, we propose a dynamic codebook updating methodology by optimizing the mutual information between the resultant codebook and relevance info, and also the workload balance among nodes that manage totally different codewords. In order to any improve retrieval performance and reduce network price, indexing pruning techniques are developed. Our comprehensive experimental results indicate that the proposed approach is scalable in evolving and distributed peer-to-peer networks, whereas achieving improved retrieval accuracy.