This paper deals with the problem of shape reconstruction from an unorganized and not necessarily uniformly distributed point cloud with outliers. A shape reconstruction method based on a Delaunay filtration called LDA-α-shapes has been introduced in 2010. This method may be used to reconstruct 2-D or 3-D shapes from a non necessary uniformly distributed point cloud but unfortunately it faces problems when there are outliers. For this reason, a little modification of the LDA-α-shapes definition is proposed in this article to deal with this kind of point clouds. The resulting new algorithm is still simple and efficient.