Abstract: A reversible information hiding method based on multiple prediction values is proposed in this paper. Multiple Predictors are firstly applied to predict a pixel value. The indexes of maximum prediction value and minimum prediction value are thus determined. Then,the relationship among maximum prediction value,minimum prediction value and current pixel value is used for classification. If the maximum prediction value is equal to the minimum prediction value,the optimal prediction value is the minimum prediction value. If the current pixel value is equal to or greater than maximum prediction value,the optimal prediction value is the maximum prediction value. If the current pixel value is equal to or less than minimum prediction value,the optimal prediction value is the minimum prediction value. Finally,the difference value between the current pixel value and the optimal prediction value is calculated and the corresponding different value is selected for data hiding. The method can extract secret information accurately and recover the original image without loss of information. The stego-image has good visual quality.

Key words: multiple prediction values, classification, difference value, reversible information hiding