A NOVEL IMAGE ZERO-WATERMARKING SCHEME BASED ON NON-UNIFORM RECTANGULAR

This paper proposes a novel time-domain zero-watermarking algorithm based on Non-Uniform Rectangular Partition (NURP). NURP is an image representation method that represents the image pixels by some bivariate polynomials over partitioned sub-areas. The partition grids and polynomials can describe the image feature properly. The host image is partitioned into different rectangle grids and the numbers of rectangles in each 8x8 region are counted and recorded as a feature matrix which will be scrambled by the Arnold scrambling method to enhance the algorithm security. The scrambled feature matrix and binary watermarking are stored as the zero-watermark. When doing the watermark extraction, the Speed-Up Robust Features (SURF) algorithm is applied to check and recover the host image if the rotation attack is detected, the rotation-recovered host image is NURP partitioned again to get the feature matrix which will be compared with the stored one and determine what the scrambled binary watermark is. The inverse Arnold scrambling method will be applied to extract the final binary watermark. Experimental results prove that this algorithm is robust to various attacks, such as JPEG compression, rotation, Gaussian noise and salt and pepper noise, etc.