• K PRABHA

• Robust color image watermarking by elliptical phase modification based on Walsh Hadamard Transform and Triangular Vertex Transform

Robustness is an essential characteristic of a watermarking algorithm. In order to achieve higher robustness, this paper proposes a novel Triangular Vertex Transform (TVT) and Walsh Hadamard Transform (WHT) based robust color image watermarking by elliptical phase modification. The existing TVT converts the RGB components of an image to U, V and W coefficients suitable for data embedding. However, embedding data directly on W coefficient of TVT using traditional embedding algorithms highly affects the quality of the imageand reduces the robustness of the watermarking algorithm. Therefore an elliptical phase modification embedding algorithm is introduced to embed the data efficiently in the W coefficient. Initially, this embedding scheme transforms the RGB color image to TVT coefficients using the TVT. The WHT coefficients are then estimated by subdividing the W coefficients into 4 9 4 subblocks. Three WHT coefficients from the first row are chosen for elliptical coordinate formation. The elliptical coordinate is formed such that the minimum and maximum values from three selected WHT coefficients form the length of the semiminor and semimajor axis of an ellipse, respectively. The phase (q) of intermediate coefficients is estimated from the ellipse, such that the intermediate coefficients act as the radius at point P. The data is embedded by modifying the phase (q) and the new radius provides the data embedded new intermediate WHT coefficient. The embedded 4 x 4 subblocks are then merged after obtaining the inverse WHT. Finally, the watermarked image is estimated by obtaining theinverse TVT. In the extraction phase, the ellipse is again formed using the same three WHT coefficients. The embedded phase q is estimated from the intermediate value, which is used for extracting the hidden data. Theproposed watermarking algorithm shows higher performance than those of the traditional watermarking algorithms; particularly, it attains optimal Peak Signal to Noise Ratio of 44.76 dB and 43.46 dB for two different embedding capacities.