Standard vs Custom Battery Packs - Decoding the Power Play
Database saliency for fast image retrieval
1. DATABASE SALIENCY FOR FAST IMAGE RETRIEVAL
ABSTRACT
The bag-of-visual-words (BoW) model is effective for representing images and videos in
many computer visionproblems, and achieves promising performance in image retrieval.
Nevertheless, the level of retrieval efficiency in a large-scaledatabase is not acceptable for
practical usage. Considering thatthe relevant images in the database of a given query are
morelikely to be distinctive than ambiguous, this paper defines “databasesaliency” as the
distinctiveness score calculated for everimage to measure its overall “saliency” in the database.
By takingadvantage of database saliency, we propose a saliency-inspired fast image retrieval
scheme, S-sim, which significantly improves efficiency while retains state-of-the-art accuracy in
image retrieval.There are two stages in S-sim: the bottom-up saliency mechanismcomputes the
database saliency value of each image by hierarchicallydecomposing a posterior probability into
local patches andvisual words, the concurrent information of visual words is thenbottom-up
propagated to estimate the distinctiveness, and thetop-down saliency mechanism discriminatively
expands the queryvia a very low-dimensional linear SVM trained on the top-rankedimages after
initial search, ranking images are then sorted ontheir distances to the decision boundary as well
as the databasesaliency values. We comprehensively evaluate S-sim on common retrieval
benchmarks, e.g., Oxford and Paris datasets. Thoroughexperiments suggest that, because of the
offline database saliencycomputation and online low-dimensional SVM, our
approachsignificantly speeds up online retrieval and outperforms the state-of-the-art BoW-based
image retrieval schemes.