於DWT-SVD-DCT轉換域所發展之影像浮水印技術

An image watermarking technique developed on the DWT-SVD-DCT domain

本文提出以離散小波轉換結合奇異值分解與離散餘弦轉換等技術之優點所建構出之新穎數位浮水印技術，其中離散小波轉換可解析低頻的影像，奇異值分解可擷取各層的幾何紋理，離散餘弦轉換則用來分析特定頻率成份。本文先試著找出適當轉換域的特徵參數，其後使用累進量化索引調變來嵌入浮水印位元，所獲結果進一步與其它類似的浮水印技術相互比較以印證成效。實驗結果顯示，所提出的方法在常見的JPEG壓縮和JPEG 2000壓縮的攻擊下，明顯表現出非常傑出的強健性，其對於一般之雜訊、濾波、旋轉還原、剪裁、亮度調整等攻擊也可有效抵抗。

This paper presents a novel image watermarking technique that integrates the virtues of discrete wavelet transform (DWT), singular value decomposition (SVD) and discrete cosine transform (DCT). The DWT can be used to acquire the low frequency image. The SVD can capture the geometric textures of various image layers, and the DCT can analyze the components of specific frequencies. This study tends to identify the feature parameters of the composite DWT-SVD-DCT domain, and then apply the progressive quantization index modulation (PQIM) rule to embed the watermark. To demonstrate the effect of the proposed scheme, the resulting performance will be further compared with other watermarking techniques. As revealed by the experimental results, the proposed scheme exhibits exceptional robustness against image compression using the JPEG or JPEG 2000 standards. This scheme can also withstand other common attacks, such as noise, filtering, rotation restoration, cropping, and brightness adjustment.