作者: Liguo Wang , Ye Zhang , Jiao Li
DOI: 10.1007/978-3-540-37258-5_87
关键词:
摘要: A new subpixel mapping method based on BP neural network is proposed to improve spatial resolution of both raw hyperspectral imagery (HSI) and its fractional image. The used train a model that describes the relationship between mixed pixel accompanied by neighbors distribution within pixel. Then can be super-resolved trained in scale. To performance, momentum employed learning algorithm local analysis adopted processing HSI. comparison experiments are conducted synthetic images truth results prove has fairly good effect very low computational complexity for HSI