A new filtering algorithm named regional growing filer is proposed in this paper. The principle of the regional growing filer is similar to the regional growing used to process digital images. First, reliable seeds are selected, then the growing rule is formulated. If the height difference between the seeds and the selected points is lower than the threshold, the selected point is regarded as the ground point, otherwise the point is removed as a feature point. When there are no laser points that can meet the rule, the growing ends. The processed point clouds use regional growing filter with no need of original data interpolation and iteration, and hence the filtering speed is fast. Experiments show that the effects of the regional growing filer is better than the results of such traditional algorithms as the maximum local slope filer and the expansion of window height threshold filer.