Abstract: This paper presents a kind of zero-watermark technique based on neural network in wavelet domain. Different from the traditional watermarking techniques, it extracts the feature from the host image and uses the feature as the watermark for copyright proof, which can eliminate the image distortion of watermark embedding. The host image is transformed via Discrete Wavelet Transform(DWT) and some coefficients at the low frequency sub-band are selected randomly as the extracted feature. A neural network is used to manage, store and detect the selected feature coefficients, which improves the practicability of the algorithm.

Key words: zero-watermark, Discrete Wavelet Transform(DWT), neural networks

摘要： 提出一种基于神经网络的小波域零水印技术。不同于传统的水印算法，零水印技术对宿主图像进行离散小波变换，并在低频子带随机选择一些系数作为提取的特征，将其作为水印用于版权证明，以避免嵌入水印导致的图像变形。通过建立神经网络模型对所选特征系数进行管理、存储和检测，以提高算法的实用性。

关键词: 零水印, 离散小波变换, 神经网络

CLC Number: 

• TP391