The shadow effect is always a major problem for optical remote sensing images classification. This effect is occurred because of topography terrain or clouds, which may cause the same material classified into two different classes. One feature of shaded area in optical images is that the values of all bands are small and this property can provide useful information to reduce the shadow effect. The band ratio approach should play an important role to eliminate the shadow effect. In this study, we proposed the adjacent band ratio approach to reduce the shaded effect and also employ an unsupervised classification algorithm. The experiment result showed that, the ratio approach certainly has potential to improve the classification performance in shaded region.