An Image Steganographic Scheme Based on Support Vector Regression
Hsien-Chu Wu, Kuo‐Ching Liu, Jun‐Dong Chang, Ching-Hui Huang · 2008
This paper presents a novel image steganographic method that utilizes support vector regression (SVR) to predict the embedded pixel value such that secret data is also embedded into the pixel-value difference between the predicted pixel value and the original pixel value. Due to the significant learning ability in the correlations of training samples by support vector regression, the trained SVR function is obtained by neighboring pixels of the sample pixels to predict the embedded pixel values, and then the proposed scheme uses pixel-value differences to embed the secret data. In the data extraction phase, the proposed scheme uses trained SVR function to predict the embedded pixel value, and the secret data is extracted from pixel-value differences. Experimental results show that SVR is good at learning the correlations of neighboring pixel, and the proposed scheme also has reliable security, high embedding capacity and better image quality for the stego-image.