一种利用影像多尺度特征的DOM不可见水印方法

数字正射影像是测绘地理信息服务的基础数据之一。根据生产单位数据现状和对遥感影像数字水印的需求，本文综合利用影像的光谱、纹理和属性特征，提出了利用影像多尺度特征的DOM不可见水印方法。该方法首先结合小波分解和Harris特征算子提取影像多尺度特征，确定多水印的最佳嵌入区域；再利用扩频化的DWT算法嵌入不可见数字水印；最后利用影像属性信息辅助提取水印。本文从主观分析、客观分析和鲁棒性分析3个方面评价了嵌入水印影像，分析结果表明，该方法能够自动、有效地在DOM上嵌入多个不可见数字水印，多水印嵌入和影像属性信息相结合提取水印的策略极大提高了水印鲁棒性和水印检测成功率，可以满足实际生产对水印不可见性、安全性、高效性和鲁棒性的要求。

Study on Invisible Watermark Method Using Multi-scale Characteristics for DOM

DOM is one of the fundamental datasets for the surveying and mapping geographic information services. According to the data status in the production unit and the application requirements of the digital watermark for remote sense images, an invisible digital watermark scheme based on the multi-scale characteristics designed for DOM is proposed in this work. Firstly, this scheme extracts multi-scale characteristics using wavelet decomposition and Harris detector. Watermark embedding areas are computed and optimized based on these multi-scale characteristics. Then digital watermarks are imbedded into these areas by using spread-spectrum DWT algo-rithm. Finally, watermarks are extracted with DOM attribute data as the auxiliary information. The watermarked DOM is evaluated in the aspects of subjective analysis, objective analysis and robustness analysis. Experiments and analysis show that multiple watermarks can be automatically and effectively embedded into DOM by the scheme proposed in this work. The strategies of multiple watermarks and attribute information greatly improve the watermark robustness and the success ratio of watermark detection. This work also proves that the scheme satisfies the invisibility, safety, effectiveness and robustness demand of digital watermark in the practical production.