Near-duplicate image detection

Detecting near-duplicate images has applications in areas such as image search, video summarization, and copyright enforcement. In this study, we propose a new algorithm for generating feature vectors for images and investigate similarity matching techniques based on the proposed feature vectors. We also construct a dataset of near-duplicate images on which to test our algorithms. In comparing results on our own dataset and a standard benchmark dataset, we find that our methods are best applied in contexts which have a narrower definition of what constitutes near-duplicate images.U of I Onlysenior thesis not approved for general distributio