A new method of building extraction from high resolution remote sensing images based on NSCT And PCNN

For more surface features can be obtained from the satellite image, it will do more benefit to the application of intensive land use and meet the requirement of building extraction with higher accuracy and more details. Considering the different scale of different type buildings in images, we proposed a new method to extract building from optical images, for example Chinese GF-2 satellite, using the Non Subsampled Contourlet Transform (NSCT) and Pulse Coupled Neural Network (PCNN). The new method not only uses the spectral, shape and texture features of the building in high resolution images and analyze the scale and the edge characterize. So it can distinguish the weak objects and can improve the buildings and some other man-made object signals. And the method combine s the NDVI and enhancement filters to eliminate the noises in order to avoid the impactions form the shadows and the plants. The experiments prove that our method could find more building areas with different scales and with high accuracy is efficient in extracting the building from the high resolution remote sensing images.