網格式空間資料的數位浮水印演算法

Abstract

本研究主要的對象為航測、衛星影像資料及數值地形模型，根據其空間資料的特性提供版權保護的方法。對於航測、衛星影像資料為以特徵點為基礎的數位浮水印演算法，結果顯示嵌入浮水印後對影像原始品質破壞很小，因此可避免空間影像資料的後續應用的成果受到影響。在特徵點的基礎下，根據特徵區域的紋理敏感值對浮水印的嵌入強度進行調整，使嵌入浮水印後的影像變形量更低，在遭受幾何攻擊時更具有強韌性。而由實驗的結果也證明，本研究所提出的影像數位浮水印演算法可以抵抗各種不同型態的影像處理和幾何轉換的攻擊，同時保持其空間資料的品質。針對空間影像常用的RST(旋轉、尺度及平移)與多項式的幾何改正對浮水印演算法進行強韌性測試，結果顯示利用RST進行幾何改正後的影像，所擷取出的浮水印正確率相當高，惟以多項式進行幾何改正後所取出的浮水印正確率則不佳。由此可知，本研究所提出的數位浮水印演算法對於多項式的影像轉換仍有改進的空間。對於數值地形模型提出以坡向及映成函數為基礎的浮水印演算法，浮水印的位元可均勻且多次的嵌入至數值地形模型的不同位置中，除了增加多餘嵌入次數，提高浮水印嵌入容量，也克服資料冗餘度較小的問題。即使數值地形模型資料因為平滑與裁切等資料處理而變動，仍能有相當好的表現，較一般的數位浮水印演算法有很大的進步。

This study mainly focused on raster spatial data for digital watermarking analysis and discussion, the explored objects included photogrammetric images, remote sensing images and digital terrain model (DTM). The watermarking algorithm of images are mainly is developed based on image's feature points for digital watermarking algorithm, and influences and robustness of images after watermark embedding were tested during the study. When conducting copyright protection using watermarking algorithm in the future, relevant results can be used as the reference when evaluating derivational problems from re-processing and geometric transformation of images with watermarks embedded. Moreover, the watermarking algorithm proposed in this study had little influence on the original quality of images after embedding, thus could avoid subsequent application outcomes of images. It was proved by the experimental results that in addition to a certain degree of robustness, the digital watermarking algorithm proposed in this study can also resist attacks from different types of image processing and geometric transformation, and at the same time maintain the quality of images. The second watermarking algorithm for image proposed in this study also based on feature points and adjusted the embedding strength of watermarks according to feature region's texture sensitivity value, resulting in less anamorphose of images after watermark embedding and higher robustness when images were under geometric attacks such as rotation and resizing. Watermarking algorithm was tested for robustness using geometric corrections such as rotation scale translation (RST) and polynomial transformation targeting remote sensing images, the NC value of watermarks captured from geometrically corrected images through RST is very high. However, the extracted watermark after polynomial transformation is unidentifiable and has a small NC value.The third part of this study is DTM, which is an important national fundamental data. As confirmed through experiments, an aspect-based algorithm was proposed in this study so that the bits of watermark information could be evenly and repeatedly embedded into DTM data, through which not only the number of redundant embedding times could be increased and watermark embedding capacity could be improved, the problem of relatively low data redundancy can also be overcome, thus the data contents would have good robustness even if they were altered because of being attacked.