True color images can not be synthetized due to some remote sensing data missing blue band. A method of simulating true color remote sensing image is proposed. The MODIS land surface reflectivily product (MOD09) is used as reference image where to select pixel samples, and a non- linear regression analysis model is acquired through a back- propagation (BP) artificial neural network that used to fit the spectral reflectance relationship among the blue band and red, or green and near-infrared bands. Then Landsat TM/MSS and SPOT- 5 blue bands are simulated with the trained fitting model, and a natural- color image is acquired. The experimental result shows that using the BP neural network model, the blue band of a multispectral image is well simulated and even the atmospheric effect of the blue band is removed. Thus a more realistic and informative natural- color image is acquired. The research results further extend the research depth on remote sensing image band simulation model based on the image-image relationship.