Research Of Image Digital Watermarking Algorithm

With the rapid growth of Internet usage, digital multimedia products (image, video, audio, text, etc.) become more and more easy to be acquired, copied and transmitted. Consequently, the copyright protection of these products turns to an important issue. Digital watermarking is just an effective technique for protecting their copyrights. The copyright information—digital watermark is embedded into digital products to verify their ownerships. The embedded watermark is usually invisible or imperceptible. However, a detector can be used to extract it or determine its existence. This technique can be traced back to as far as early 1990's. Since then, it has drawn extensive attention from both academic and commercial communities. Digital watermarking has become one of the domestic and foreign hot issues of research. Though a lot of watermarking techniques have been developed, digital watermarking is still in his infancy, and many problems remain yet to be solved. This paper addresses research of image digital watermarking algorithm. The main contributions are summarized as follows: 1) Review the research background and practical value of digital watermarking, introduce its fundamental frame, features and application fields, sum up the current developing condition, and point out some problems of digital watermarking. Then common mathematic knowledge (fourier transform, cosine transform, wavelet transform) is deeply analyzed. Some classical algorithms are surveyed. 2) Referring the idea of JPEG compression, making full use of human visual system and local characteristics of image, the author puts forward a novel adaptive blind digital watermarking scheme based on DCT (discrete cosine transform). Experimental results show that the embedded watermark is invisible; while at the same time, the algorithm is robust to some typical kinds of image processing such as scaling, salt & pepper noise, filtering, especially, JPEG compression and cropping attacks. Watermarking detection doesn't need original image. It is a blind detection. 3) Consulting the idea of EZW (the embedded zerotree wavelet algorithm), the author proposes a new wavelet-tree digital watermarking algorithm based on DWT (discrete wavelet transform). First, decompose the image at level three utilizing db1 wavelet basis; then search the significant coefficients in the low-frequency and the wavelet tree of significant coefficients in the mid-frequency; finally embed watermark into these coefficients by the formula of addition. The intensity of watermark is determined by scaling parameter. Experimental results indicate that embedding partial watermarks into the significant coefficients of low-frequency and other partial watermarks into those of mid-frequency can not only insure the robustness but also the quality of watermarking image changes little and human can't almost perceive the change of image. The scheme can resist against some common image attacks such as JPEG compression, median filtering, Gaussian lowpass filtering, salt & pepper noise, etc. Particularly, the method is well robust against JPEG compression.