Image authentication using Zernike moment watermarking

The rapid development of internet and digital image editing softwares facilitated the access and illegal usage of digital images. Digital watermarking emerged as a unique tool to protect the authenticity of the image. It is a technique of insertion of a message into a cover media imperceptibly. In this thesis, the main objective is to design and develop a watermarking system for image authentication. A semi-fragile watermarking system using content-based techniques is proposed in this work to solve the problems with the previously proposed methods which are not robust against normal image processing operations such as rotation, scale, translation, noise addition and JPEG compression. The Zernike moments and Sobel edge map are used as watermarks to provide malicious attack classification and locating tampered area. The Zernike moments are chosen due to special characteristics such as robustness against rotation, scale, translation (RST) and noise. Also an image can be reconstructed from the extracted Zernike moments. These watermarks are generated from the approximation component of discrete wavelet decomposition in third level. The Sobel edge map is fused in wavelet coefficients of the corresponding component, and Zernike moments vector is fused in Sobel edge point of the original image. A comparison between recovered Zernike moments and generated Zernike moments form watermarked image at receiver determine the authenticity of the image. In case of authenticity failure, a comparison between recovered Sobel edge map and recently obtained Sobel edge map from the watermarked image detect the location of tampered areas. The results from the test show that the proposed algorithm has good performance in discriminating the malicious from non-malicious modifications which is the main character of an image authentication algorithm. This work significantly improves the robustness of semi-fragile characteristics of watermarking. It accepts scale, translation, noise pollution, rotation and JPEG compression modifications on the watermarked images,so it can solve the problem with previous methods with flaw of discrimination between malicious and non-malicious attacks. The adding or replacing a portion of the image is regarded as malicious attacks and rejected by this algorithm. Using two different watermarks lead to a good classification of incidental and malicious modifications and locating the tampered areas. Experimental results showed that this algorithm achieves better bit error rate against JPEG compression, rotation, scale and noise attacks in comparison with other reviewed schemes in this work. A great improvement is also achieved in the capacity of the watermarking system as shown in chapter Four. This scheme also offers better classification of malicious modifications over the other reviewed schemes in this work.

Preview PDF FK 2013 79IR.pdf Download (1MB) | Preview

Actions (login required)

View Item