Compare Between DCT and DWT for Digital Watermarking in Color Image

摘要

In This paper we compare between Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT) in the field of image authentication and digital watermarking. Our proposed method depending on the embedding stage and extraction stag that presented in [1] but our scheme embeds the logo bits inside the low frequency domain because DWT gives optimal results with LL domain, while [1] used middle frequency domain with DCT. Our improvement by using a secrete K called user key (int. K=1, K<=N) where N=10, is used to generate a random vectors for the selected coefficient, this attempt is made to increase the security and robustness for the proposed scheme to compare between DCT that illustrated in [1] with middle frequency domain and DCT, DWT in the existing paper that used low frequency domain, it used block based technique with embedding stage, the development that illustrated in the existing paper by using DWT transform on the recovered of binary watermark for the purpose of image authentication in frequency domain using DWT, and DCT transforms with true color image. The method focused on the objective quality after embedding stage and the recovered watermark after extraction stage. With DWT in the first step, the cover image is decomposed into three levels by DWT transform. Then the hiding site was LL sub band of the DWT coefficients. Furthermore, our proposed method deal with true color image without converting its color space into other color space with various image texture all of them with size of 256*256 Bit map image file format, the proposed scheme deal with three sub-bands (Red, Green, and Blue) at the same time to hide logo bits inside the host by using Patchwork technique with embedding stage, so if one is destroyed the other may survive, it provide optimal security whenever any sub-bands color destroyed. With our proposed method a secret watermark in the form of binary (o, 1) pattern is embedded inside the host under DCT, DWT, one bit from the watermark will be embedded inside the selected coefficient from the selected block of the host. Our proposed method was evaluated with different types of intended attacks such as: salt and pepper noise, Poisson noise, and speckle noise. Moreover, unintended attacks consider by spatial enhancement filter such as median filter that used to improve the quality for the watermarked image after unintended attack. After experiments, it was found that our proposed method provides security and high performance with low computational complexity and good objective quality. Our scheme evaluate the imperceptibility for the watermarked image after embedding stage by using Peak signal to Noise Ratio (PSNR), while the recovered watermark evaluated by some types of metrics such as Mean Square Error (MSE), Normalized Correlation (NC), and correlation factor (SIM). Our proposed method has ability to deal with different image texture and format such as (BMP), Portable Network Graphics (PNG), and Tagged Image File Format (TIFF). Keywords: Authentication, Objective, Subjective.