Robust Image Watermarking Using Adaptive Structure Based Wavelet Tree Quantization

Abstract

This work presents a novel robust wavelet-tree-based watermarking method based on structure-based quantization. Wavelet-trees are arranged into super-trees. The watermark bits are then embedded into the super-trees by using the proposed structure-based quantization method. Next, the super-trees are quantized into a significant structure according to these bits. The quantized super-tree has a stronger statistical characteristic than the unquantized super-tree. Based on this characteristic, the watermark bits could be extracted robustly after an image distortion attack. Finally, an adaptive method is developed to raise the PSNR value. Compared with Wang et al. [17] method, the proposed adaptive method increases PSNR about 5.83 dB. The proposed method also has a higher maximum number of watermark bits than other methods, thus increasing the capacity for embedding. Besides, its computation load is low. Experimental results demonstrate that the proposed watermarking method using adaptive structure-based wavelet-tree quantization performs well in JPEG compression, filtering (Gaussian filter, median filter and sharpen) and geometric attacks (pixel shifting and rotation). In addition, it is very robust against multiple watermark attacks.