In this paper we proposed a novel supervised dimensionality reduction method, named Discriminant Isometric projection. The aim is to compact the data points from the same cluster on high-dimension manifold to make them closer in the low-dimension space, and to make the ones from the different cluster further, which is beneficial to preserve the homogeneous characteristics for classification. We compared our method with other three methods for dimensionality reduction over the ORL face dataset and experiments show that Discriminant Isometric projection produces stable performance and good precision. © 2009 IEEE.