Research On Digital Image Watermarking Based On Integer Wavelet Transform

As a useful means for the copyright protection, watermarking technology has gained more and more concerns since it appeared in 1993.The research of this thesis is image watermarking technique. At present, most of the image watermarking algorithms based on the wavelet transform fail to take into account the shortcoming that the first generation wavelet transform used in image processing will introduce quantization errors inevitably, because of the storing characteristics of the image. Wavelet transform that maps integers to integers will not introduce quantization errors. Its time-frequency analysis method and multi-scale property accordant with human being vision system makes integer wavelet transform an ideal way for digital watermarking. This thesis presents two novel digital image watermark algorithms based on the integer wavelet transform.A semi-fragile watermarking approach, which is based on integer wavelet transform and capable of tamper location, is proposed. With this method, we can insert content-based watermark, and we have applied scrambling transformation and chaotic encryption, which can resist both statistics attacks and copy attacks. After the host image was DWT-decomposed, we choose the block of intermediate frequency coefficients as the location for watermark insertion, which takes full advantage of correlation of neighboring coefficients. Experimental results demonstrate that the algorithm is imperceptible, blind, robust to JPEG compression, sensitive to malicious tampers such as cut and cut-replace, and precise to tamper localization. Moreover, at present the majority digital watermarking system inserts the fixed watermark information sequence into the image, therefore this kind of watermarking system has two shortcomings: one is that the verification process needs to know the original sequence, and this kind of sequence transmission process itself has enlarged the authentication risk; the other is that this algorithm is liable to forge attacks. This algorithm has overcome the above shortcomings effectively. This algorithm unifies"the integer wavelet transform"and"content-based watermark"effectively, and has obtained the good effect.This thesis proposes a lossless digital watermarking algorithm based on Harr wavelet transform, which can recover the original image without any distortion from the marked image after the watermark has been extracted. The selection of wavelet coefficient based on Harr wavelet inverse transformation is to prevent grayscales from possible overflow/underflow of 0~255 that may take place due to the data embedding. This algorithm hides the watermark into coefficients in high frequency sub-band of Blue color component. Experimental results show that it can embed much more data compared with the existingreversible data hiding techniques. The imperceptibility requirement is satisfied and blind extraction of the embedded watermark data is also achieved.