![Sample our Geography journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5937/TOC-Subject-Sample-2021-Geography.jpg"})
Abstract
Super resolution (SR)-based spectral unmixing (SRSU) is a recently developed spectral unmixing approach. Some issues in SRSU, such as the role of training database, remain unclear. According to the example-based SR reconstruction, training databases impact the reconstruction and consequently may also impact SRSU. This study investigated how training databases affect the SRSU, so the training database can be designed appropriately. A Markov network model was employed as a way to implement SR reconstruction. Ten training databases, derived from different types of remotely sensed images and non-remotely sensed images, were constructed and each training database-based SRSU was evaluated with respect to unmixing accuracy. Surprisingly, the experiments revealed that the SRSU performance is not sensitive to the types of training databases, since different training databases resulted in similar accuracy in spectral unmixing.
Acknowledgements
We are grateful to Prof. Tim Warner from West Virginia University and two anonymous referees for their constructive comments. This research was supported by the National Nature Science Foundation of China under grant nos. 41101413 and 41023001, National Basic Research Program of China under grant no. 2011CB707101, Ph.D. Programs Foundation of Ministry of Education of China under grant no. 20110141120073 and the Scientific Research Foundation of Key Laboratory for Land Environment and Disaster Monitoring of SBSM under grant no. LEDM2011B02.