A NEW METHOD FOR DIGITAL IMAGE WATERMARKING BASED ON VECTOR QUANTIZATION (VQ) Mohamed A. Abdel-Wahab, High Institute of Energy, south Valley University, Egypt Hany Selim and Electrical engineering Department, Faculty of engineering, Assiut University, Egypt. Usama Sayed. Electrical engineering Department, Faculty of engineering, Assiut University, Egypt. Email: Usama@aun.edu.e g Abstract: Applying vector quantization (VQ) to an image delivers an index map. Dividing this index map into blocks results in each block using a limited number of codeword indices of the whole codebook (CB) while the rest of indices are unused. The proposed method exploits the unused indices to embed the watermark bits. The proposed watermark gives a high peak signal to noise ratio (PSNR) while keeping bit per pixel (bpp) at a small value, so it can be easily used in transmitting digital images over the internet. Furthermore, the reconstructed image has robustness against different attacks such as cropping, JPEG compression, and median filter. So it can be used in copyright protection. KEYWORDS: Image watermarking, vector quantization, index remapping I. INTRODUCTION With the widespread use of the internet and the rapid development of multimedia, the need for copyright protection became very important. A variety of image watermarking methods has been proposed, where most of them are based on the spatial domain [1, 2] or the transform domain [3, 4]. However, in the recent years, several image watermarking based on VQ domain appeared, since these schemes can enhance the traditional VQ system by adding the watermarking ability. Image watermarking used for copyright protection must have the following properties: 1- Imperceptibility: after the embedding process, the eye must not distinguish the difference between watermarked image and the original image. 2- Robustness: applying image processing such as JPEG compression, sharpening, blurring to the watermarked image, should not seriously affect the embedded watermark. A. Vector Quantization VQ is a lossy compression method, it is used for image compression due to its high compression ratio and simplicity [5]. First, the image is partitioned into non- overlapping blocks of r × u pixels. Each block is converted into a vector with dimension 1 × k (where k = r×u). After that, each vector x is mapped into its nearest codeword c of the codebook using the minimum Euclidean distance. If the codebook contains N codewords then after mapping all vectors in the image by their ١