FEATURE POINT EXTRACTION BASED GEOMETRICALLY RESILIANT DIGITAL IMAGE WATERMARKING

Abstract

A digital image watermarking scheme based on the features of the original image has been implemented in order to resist geometrical transformations. Scale invariant feature transform(SIFT) has been used for interest point extraction in order to locate the invariant regions for signal embedding. Additionally, the watermarked image was subjected to various image processing attacks such as blurring, rotation, shifting, addition of noise (Gaussian, salt & pepper), histogram equalization, JPEG Compression etc. in order to damage the inserted watermark and produce copyright infringement issues. This thesis work reviews the performance of various watermarking algorithms in the presence of different types of attacks to justify the robustness of the algorithm. The parameters considered to review the performance under these attacks were Peak Signal to Noise Ratio(PSNR), Normalized Correlation(NC) and Tamper Assessment Function(TAF). The performance of this algorithm has been compared with known conventional image watermarking techniques such as Discrete Cosine Transform(DCT), Discreet Wavelet Transform(DWT) and DWT-SVD (DWT Singular Value Decomposition) and verified using MATLAB.