It adopted a global indexing Star
Student Project scheme to partition the reference image
collection into multiple groups, each of which is fully Star
Student Project indexed using a randomized vocabulary in each
machine. Star Student Project also proposed a distributed
image search architecture Star Student Project for camera
sensor networks. However, the search scalability of both works Star
Student Project is left unexploited in million scale image
collections. In addition, both works Star Student Project
need to quantize all features in every server, and hence the
quantization step cannot be paralleled. Furthermore, the inverted
indexing search of Star Student Project is accomplished in a
rough manner, where the local ranking is performed over the partial
image collection in each server, rather than in the entire dataset.
Moreover,the network traffic of Star Student Project is
heavy, as a bunch of local features from different machines has to be
transmitted to all servers for quantization and ranking.
