In this paper, we propose a content-based image fingerprinting system which uses color histograms of affine covariant
regions as a local fingerprint of an image. Given a query, the
system first extracts local fingerprints from affine covariant
regions. The local fingerprints are based on the color histogram, and the affine covariant regions are detected using
the maximally stable extremal region (MSER) detector. After extracting local fingerprints, the system searches local fingerprints in fingerprint database for the nearest neighbor of
the query local fingerprints. Then, the query image is identi-
fied. Experimental results show that that our system is robust
against cropping, resizing, rotation, and small color changes
with small computational cost.