Summer Training Project Report on “Digital Watermarking based on fractal image coding using Ridgelet Transform and DWT”

We propose a newmethod for digital watermarking of images based on Fractal Image Coding using Ridgelet Transform and Discrete wavelet transform (DWT). The method takes Ridgelet transform of the image to get a transformed image and performs 2-level Discrete Wavelet Transform(DWT) on the transformed image to get low frequency sub-bands LL2 and LL1. LL1 is used for range blocks and LL2 is treated as domain pool for domain blocks.We embed(implant) the watermark into LL1 based on fractal image coding by replacing LL1 range blocks with transformed domain blocks from LL2 sub-band. We exploit the mutual similarities between the LL1 and LL2 sub-bands for this step. The watermarked image is of very good quality because we first perform Ridgelet Transform on the original image and then apply DWT and watermark embedding procedure. This preserves the quality of the original image to a great extent. Despite embedding the watermark into LL1, the watermarked image quality is not degraded to extremities. However, this gave us benefit in that the watermarked image is less prone to JPEG compression. This method is very successful for embedding the watermark for limited number of times only, after that the degradation becomes quite high. Results reinforce that, compared to existing conventional watermarking techniques our proposed method improves robustness and maintains quality. For a binary watermark of length 32 bits, PSNR is as high as 55 for embedding the watermark only once. For embedding 4 times the PSNR is 43. The conclusion based on the method is that applying watermark to low frequency sub-band LL1 of ridgelet transformation of an image gives very high qualitywatermarked image and because we embed the watermark in LL1, it is less prone to JPEG compression. The method performs well as seen in the results.