Fast retrieval of high-dimensional feature vectors in P2P networks using compact peer data summaries

Abstract

The retrieval facilities of most Peer-to-Peer (P2P) systems are limited to queries based on a unique identifier or a small set of keywords. The techniques used for this purpose are hardly applicable for content-based image retrieval (CBIR) in a P2P network. Furthermore, we will argue that the curse of dimensionality and the high communication overhead prevent the adaptation of multidimensional search trees or fast sequential scan techniques for P2P CBIR. In the present paper we will propose two compact data representations which can be distributed in a P2P network and used as the basis for a source selection. This allows to communicate only with a small fraction of all peers during query processing without deteriorating the result quality significantly. We will also present experimental results confirming our approach.

Aspects of Adaptivity in P2P Information Retrieval

“…1 we show two experiments performed using the source selection based method described in [15] and using PRISM. By changing the distance measure at query time, we also simulated a relevance feedback query.…”

“…This way, the 50'000 images were distributed unevenly over ≈2'600 peers. From all images in both collections we extracted 166-D color histograms using 18 × 3 × 3 HSV color bins and 4 grey levels, as described in [15].…”

“…These summaries can be used for collections of images and have been tested on 166-dimensional histograms extracted from stock photos and consumer photos [17,15].…”

“…As shown in [15], such a clustering can be calculated efficiently, i.e. without having peers transfer their data collection.…”

“…The simplest method is the one described in [17] and will be presented here: When processing queries, the query peer first finds which center c i is closest to q. We call the closest center c i q .…”