Research On Application Of Fractal In Image Watermarking And Secret Communication

Digital image watermarking technique is one of the investigation hotspots in the image academic field in the world. This thesis studies the digital image watermarking algorithms based on fractal coding, gives the classification of the existing fractal-based image watermarking algorithms. A new image watermarking algorithm is proposed, a novel secret communication scheme is devised based on the mutual similarity of images and some relevant results are given in this thesis.There are three kinds of image watermarking algorithms based on fractal in the existing approaches. The first method inserts the watermark by changing the fractal coding parameters, the second utilizing the image self-similarity, and the last kind combining fractal and the other theories. This paper presents a novel watermarking algorithm based on the fractal coding. The method changes the search range of the domain block and geometry transform simultaneously. It can insert one watermark as well as two different watermarks. The simulation results show that the one-watermark-inserted algorithm can improve the reliability of the watermark extraction when the watermarked image is attacked, especially by cropping. The two-watermark-inserted algorithm extends its application area and can be used to protect copyrights and follow the illegal dispense of legal users. No original image is required for watermarking extraction.The secret communication is one of the important problems studied in the communication field. This thesis proposes a new secret communication way using the image mutual similarity. The transmitting and receiving ends need to posses a common reference image database. The transmitting end use the reference image as the search range of the domain block, and get a group of fractal codes by comparing the transmitted and the reference image. Only the fractal codes need be transmitted in the communication channel and the receiving end iterates the fractal codes in the reference image to get the transmitting image. The simulation results show that even if the transmitting image and the reference image are not resembled, a good result can be obtained by one iteration in the reference image. Also, this method can be used to compress the image data.