Tattoo image retrieval for region of interest
Joonsoo Kim, He Li, Jiaju Yue, Edward J. Delp · 2016
In this paper we present our submission to the NIST Tattoo Recognition Technology — Challenge (Tatt-C) that was held in early 2015. The goal of our work is to match a subregion of an tattoo image to another image. We introduce an image descriptor that combines SIFT (Scale Invariant Feature Transformation) and LSS (local self similarity). The SIFT descriptor uses the histogram of the pixel gradient and the LSS descriptor uses local similar color patterns. We introduce a weighted distance similarity metric for image retrieval. By using image similarity based on the combination of a sigmoid function and the distance between the image descriptors we improve image retrieval accuracy. Experiment results show that our method achieves good performance compared with several other methods.