We evaluated the resistance of image compression for the zero-watermarking-type steganography which proposed by us in 2005. As a result, even if we compressed to 1/100 of an original image, it still can extract its watermark information with an accuracy of 99 % or more. We also proposed a new type of sparse filtering for low-frequency spatial extraction, and confirmed its performance by computational experiments. Finally, we mentioned a study on the relevance of the proposed filter and biological visual systems.
展开▼
