A Print-Scan Modeling Scheme Based on BP Neural Network

Aims at the complexity of print-scan process, this paper presents a scheme to construct print-scan model using BP (Back Propagation) Neural Network principles. In this scheme, print-scan process is treated as a black-box. Firstly, divide RGB color space into several color blocks, get the color mapping table between the original colors and that after PS operation. Then obtain the trained net model by utilizing the BP Neural Network algorithm. At last, simulated PS (print-scan) image can be gained from the trained model by taking the original image as the input. Experimental results indicate the modeled image is quite close to the real PS one, which proves the great feasibility of this modeling scheme.