Recent advances in the sensors technology of remote sensing have led to an increased availability of multi-sensor data from the same area.In particular,hyperspectral(HS)images provide a detailed description of the spectral signatures of ground covers,whereas Light Detection And Ranging(LiDAR)data gives detailed information about the height of the same surveyed area.When using these data,one should consider the following important issues on data acquisition and applications.For data acquisition,despite advances in sensor technology,HS images are inevitably degraded by noise and blur,which can affect information retrieval and content interpretation.Existing methods perform image restoration directly on the original HS data.However,in many real world applications,the high dimensionality of HS data as well as the redundancy(high correlation)between the bands,make the processing of HS images very computationally intensive.For application(e.g.classification),it is clear that no single technology can be sufficient for a reliable classification.The stacked architecture of data fusion(simply concatenate several kinds of features together)has been widely used for the classification of multi-sensor data due to its simplicity.However,this method does not take into account the properties of different data sources,as a consequence,result in the risk of worse performances than even using single feature.