1. Study the paper “Image watermarking based on DWT coefficients modification for social networking services” and use the coded provided to conduct the experiment. The main file is “Method_H25fW25f_Bit_256.m”.
2. Transform the watermark binary “panda” (panda.bmp) using 2D-DWT. Use the coefficients in the LL sub-band of the transformed logo image as a watermark. Repeat the same steps with the host image “Lena” (Lena.bmp).
3. Embed the watermark into the LL sub-band of the transformed host image, and adjust the signal strength until the PSNR value of watermarked image reaches 30±0.1dB. Note that the PSNR measurement must be performed on the watermarked image in spatial domain, i.e. the embedded watermarked coefficients in the LL sub-band together with the intact coefficients in the remaining three sub-bands must be invert transformed to reconstruct the practical watermarked image.
4. Extract the embedded watermark using the mean filter based prediction techniques (function “Predict_Mean_Filter”). Compress the watermarked image with the JPEG standard at 90 and 80 quality levels, and extract the embedded watermark using the same previous technique.
5. Record the values of PSNR and NC at different versions of watermarked images and extracted watermarks.
6. Repeat steps 1-5 using the HL, LH and HH sub-bands as the working domain, respectively.
7. Finally, discuss about the results you obtained. Explain the relationship between different variables, including the accuracy of extracted watermark from different sub-bands.