Image Self-Embedding Algorithm For Content Reconstruction

This thesis first describes the research background and significance of the self-embedding watermarking technology, and then introduces the basic knowledge, performance evaluation and the domestic and foreign research status of self-embedding watermarking algorithm. The second chapter of thesis focuses on the existing self-embedding algorithm using fountain code, and deep analysis and simulation are carried out on these algorithms.At the same time, the problems of existing algorithms are pointed out. In this thesis, two new self-embedding algorithms are designed to solve these problems.This thesis first designed a self-embedding algorithm based on gradient feature and self-adjustment mechanism. The steps are as follows:according to the defined gradient feature, image blocks are divided into different types; according to the designed self-embedding algorithm, basing on "the ideal load of watermark types", the embedding capacity of weigh type watermark and recovery watermark can be adaptively calculated by "proximity principle "; calculate the target fault-tolerant rate and the amount of desired recovery information of the test image; according to the size relationship between original-information and expect-information, increase or decrease the recovery information of test image, at the same time, update the reconstruction template to adjust the recovery information, which is used in object fault tolerance fitting; finally, Generate watermark by encoding the classification information and recovery information by fountain code, and embed them into the image. Compared with existing algorithms, the designed algorithm has following advantages:(1) The fault-tolerant rate of all image can reach a maximum value above 70%, at the same time, the invisibility is around 44.1dB. (2) Our algorithm can balance the fault rate and reconstruction quality adaptively, and needn’t set the target tolerance rate in advance, as well as recording the "significant areas". Besides, our algorithm reduces the subjectivity of algorithm. (3) Our algorithm achieves a best fitting theory of the fault rate of type information and recovery information fault rate, which can relieve "redundant embedding" problem.In order to improve the reconstruction quality of complex image, enhance the fault tolerance under random tampering pattern to avoid the low local tolerance problem, a grouping self-embedding algorithm with balanced fault tolerance is designed. To ensure that the algorithm has balanced fault tolerance ability and to improve the reconstructed quality of complex contain, our algorithm first designs an image grouping compression method based on pixel-interval estimation to generate three complementary groups of recovery data. Then the unequal error protection mechanism is introduced in watermark generation for recovery:use a fountain code to encode the group of data which contains the most basic characteristic of image to generate watermark information, with the minimum rate. And fountain coding the rest two groups of recovery data to generate watermark, with a larger rate, which can increase the restore ability under different tampering rate or data loss rate and can fit different transmission conditions.Finally, this thesis designs the algorithm simulation system in order to simulate and validate the proposed watermarking algorithm with self-recovery capability.