嵌入浮水印於可調式視訊

Abstract

這篇論文是以可調式視訊編碼 (scalable video coding) 的演算法為基礎，在可調式視訊中嵌入浮水印，首先介紹可調式視訊編碼的演算法，可調式視訊編碼包含時間維度 (Temporal domain) 的轉換、空間維度 (Spatial domain) 的轉換、壓縮編碼 (Entropy coding)、和移動評估 (Motion Estimation)。移動評估是用來選擇移動向量 (Motion Vector) 的編碼模式。而我嵌入浮水印的演算法，是將浮水印嵌入在空間維度轉換後 L frame 的中頻，採用區塊的方法來嵌入，並加上安全密碼 (Security Key) 和量化區間係數 (quantization coefficient)，來加強浮水印的強固性。我們也針對不同的解析度、傳輸速率和量化區間係數做實驗，由實驗中可以看出來在嵌入浮水印演算法中，不同的解析度對浮水印的偵測會產生出相同的結果，而不同的量化區間係數對視覺感受浮水印的偵測會有不同的結果，量化區間係數愈大 BCR (Bit Correct Rate) 就愈高，但在 PSNR (Peak Signal to Noise Ratio) 上就愈差。我也使用影格丟棄 (frame drop)、影格平均 (frame average) 和移除平行線 (remove line) 三種破壞方式，企圖破壞浮水印，利用本論文所提的演算法還是可以偵測出所嵌入的浮水印。

This paper is based on scalable video coding algorithm to embedding watermark in scalable video, first we will introduce on scalable video coding algorithm. Scalable video coding algorithm include temporal domain transformation, spatial domain transformation, entropy coding, and motion estimation. Motion estimation is used to select motion vector coding mode. In my embedding watermark algorithm, watermark will embed in L frame’s middle frequency after spatial domain transformation. The concept is to use block to embed watermark and append security key and quantization coefficient to make watermark robust. We also use different resolutions, rates and quantization coefficients to do some experiments. In this experiment, we can see my algorithm to detect watermark in different resolution has same result. But different quantization coefficients will get different results in detecting watermark, so quantization coefficient increase the BCR (Bit Correct Rate) will increase, but PSNR will decrease. I also try to destroy watermark use three methods frames drop, frames average and remove line, and use this paper algorithm we can detect embedded watermark successful.