In solving intra-voxel fiber crossing problem in white matter fiber tracking, the classification of crossing and non-crossing regions seems essential and challenging. Although high C<inf>p</inf>value is a usable indicator of intra-voxel orientational heterogeneity, only using this metric is not accurate enough to decide exactly whether fiber crossing occurs as the lack of an exact and recognized threshold. In this paper, we present a new method which directly uses the orientation difference and the number of fibers that coexist in a single voxel as defining criterions to classify crossing and non-crossing voxels. Experiments with real DT-MRI data show higher accuracy and better adaptability with our new method.