To extract quantitative information from the remote sensed imagery accurately

atmospheric correction that decoupled the surface spectral signature and the aerosol scattering effect is a necessary step. There are several models to describe the multiple bouncing between the surface and atmosphere. But when viewing only from a single direction

we can’t get the enough information such as the albedo of surface to decouple the surface signature and aerosol scattering from these models

so a lambert surface has to be assumed and large error may be introduced in the atmospheric correction when the ground is non-lambertian. This paper describes the atmospheric correction algorithm of AMTIS for the VIS/NIR bands. The algorithm is based on a BRDF Loop approach and MODTRAN4.1 is applied to calculate the atmosphere parameters used in the atmospheric correction. The parameters used to decouple the multi-bouncing effect such as direct-hemisphere albedo and hemisphere-direct albedo are estimated in the loop process. The BRDF model for calculating albedo is a kernel driven model. To accelerate the atmospheric correction

all the atmospheric parameters used in this algorithm are pre-computed with MODTRAN4.1 and saved in the Look-up Table. Pri-or knowledge is used in the retrieval of kernel-driven model when the prior multi-angle sampling is not enough. The atmospheric correction result is validated using the synchronous ground experiment. The atmospheric correction result shows that this algorithm can wipe off the atmospheric smoothing effect and the BRDF shape of ground surface is reversed effectively.