An Image Hided-Data Detection Method Combining Markov Chain and Support Vector Machines

摘要

An image hided-data detection method is proposed combining 2-D Markov chain model and Support Vector Machines (SVM) by the paper, in which image pixels are predicted with their neighboring pixels, and the prediction-error image is generated by subtracting the prediction value from the pixel value. Support vector machines are utilized as classifier. As embedding data rate being 0.1 bpp, experimental investigation utilizing spread spectrum (SS) and a Quantization Index Modulation (QIM) method data hiding method respectively, correction detection rates are all above 90%. For optimum LSB method, the method achieves a detection rate from 50% to 90% above with 0.01bpp-0.3bpp various embedding data rates.