Watermark image recognition from local scale-Invariant feature

Abstract

This paper explains algorithm of digital image watermarking using DWT for copyright protection. The detail coefficients of second level host image after taking DWT is used as a watermark key for the watermark logo insertion and extraction process. A watermark image recognition system has been developed that uses new class of local image features. These local features of image are invariant to rotation, scaling and are shown to provide robust matching across a substantial range of affine distortion, change in 3D viewpoint, addition of noise, and change in illumination. The recognition proceeds by matching individual features to a database of features from known watermark images using a fast nearestneighbor algorithm, followed by Hough transform to identify clusters belonging to a single object. The effectiveness of local scale invariant feature is tested on original watermark logo, extracted watermark logo. We were able to design a prototype system, which provides user authentication. © 2009 IEEE.