| Summary: | 碩士 === 國立成功大學 === 電機工程學系碩博士班 === 92 === Because the rapid development of Internet, digitized content can be accessed, duplicated and altered easily, therefore intellectual property (IP) of Internet became more and more important. Digital watermarking is one kind of methods for protecting intelligence proprietary security.
In this thesis, a watermarking system based on Discrete Wavelet Transform (DWT) is proposed. To begin with, we transform the original image (Cover image) and the watermark image (Logo image) from the spatial domain to the frequency domain by using 1-scale and 2-scale equal band DWT respectively. After the transformation, we use SVD to choose the feature of watermark, which is referred to as the watermark element (feature). The watermark feature is then quantized and encoded ((7,4) Hamming code) to form a watermark code. Each watermark code is then spreaded by 5 bits PN code to form a watermark PN sequence. To combine (7,4) Hamming code and PN code, our scheme has ability of double error correction. For security, we rearrange blocks of the original image to break down the relationship of the coefficients in each subband, and take absolute value, sort and divide into 7 parts in each block. Finally, we used the genetic algorithm to choose a fine group coefficients of the original image adaptively for embedding the watermark.
Experimental results show that the proposed scheme can obtain good quality extracted watermarks under various attacks. The PSNR values of the watermarked image of Lena and Texture are more than 25dB.
|
|---|
