The Research Of Semi-fragile Watermarking Technology Based On Edge Detection

In recent years, with the linear growth of Internet users, Network has become an important platform for study, work and communication, Digital image, audio and video multimedia products are being applied more and more widely in the internet. But, the digital products are vulnerable to be copied, modified, transmitted and damaged in the network transmission process. This has seriously infringed the interests of intellectual property owners of digital products. So the copyright protection of digital products is urgent. Randomness and redundant data are ubiquity in digital works, digital watermark technology is to utilize these characteristics, embedding the copyright information such as digital, sequence number, characters, image into the digital works, thus play function such as copyright protection, secret communication, true and false identification of the data files and products identification.Semi-fragile watermarking technology is a branch of the digital watermarking technology divided by the robustness, which is rapidly develop in recent years. It have been putting forward a series of digital watermarking algorithms based on spatial domain, transform domain, compressed domain, statistics and physiological mode. However, Compared with robust and fragile watermark, the research of semi-fragile digital watermarking is lagged behind others.This article first introduced the development situation of semi-fragile digital watermarking technology, analyzed it's basic features, applications field, classification, basic principles, evaluation methods and conventional type of attack, pointed out the advantages and disadvantages of existing semi-fragile watermarking, and compared edge detection results with different edge detection operator. Then, a semi-fragile watermarking algorithm based on edge detection is proposed in this paper. According to brightness mask features reflected in low frequency coefficients after wavelet transform and texture characteristics reflected in edge points, the entire image will be block processed when watermark information is embedded. Different low frequency coefficients are dynamically quantified with different quantitative step. It makes the extracted watermark information keeps invariant when its element value changed in robust region. Edge-detection algorithm is that firstly determine two adaptive thresholds based on the coefficient features of image; by means of Laplacian operator's characteristics of high precision of segmentation positioning and good effects on edge detection to isolated points, secondly use one of the threshold to determine the number of the most significant edge points; thirdly compare the another threshold with the maximum value in the field except the central point of the field calculate and judge whether the pixel point of maximun value is edge point; if the pixel point maximun is edge point,then rank the pixel point as the center and judge it's field whether exist edge point. In this way, take edge tracking point by point to acqure all edge points of the whole image. After a simple random scrambling, watermark information were embedded into low-frequency coefficients, to ensure the algorithm has a strong robustness to these attacks like salt and pepper noise, image cut and so on. A new kind of adjustment operator was brought up, which not only keeps the consistency of the coefficients sum between after adjustment and before adjustment, but also make the fractional part of coefficients after inverse wavelet transform adjusted to the least sensitive pixel point to human eyes. Watermark detection not required original video, it realized the quasi-blind watermark detection.