A Sturdy Nonlinear Hyperspectral Unmixing

Abstract

Hyperspectral unmixing (HSU) is a way to process the prediction of the existing endmembers and the fractional abundances (FA) available in all pixels in the hyperspectral images. However, in a practical scenario, hyperspectral image is frequently corrupted due to many types of noises at the time of acquiring phenomenon such as dead-lines, impulse noise (IN), Gaussian noise (GN), and stripes. This type of complicated noise leads to mitigation in the quality of the acquired HSI, by making them to lose the precision process. To address these issues, this article presents a sturdy nonnegative matrix factorization (S-NMF) with integrated fast dissociable non-local Euclidean median and iterative block coordinate descent algorithm (IFD-NLEM-IBCDA), where FD-NLEM eliminates mixed noise from acquired HSI without degrading the original quality, then IB-CDA is utilized to unmix the HSI by solving S-NMF minimization problem. Furthermore, we also provided a solution for a hyper-parameter that is utilized in IB-CDA during the S-NMF minimization. Extensive simulation results, together on real and synthetic HSI, exhibit the superiority of anticipated unmixing under mixed noise conditions over conventional unmixing algorithms.

Keywords:

• Hyperspectral images
• Linear mixture models
• Nonlinear mixture models
• Nonlinear spectral unmixing
• Spectral unmixing

Additional information

Notes on contributors

M. Venkata Sireesha

M V Sireesha received her BTech degree in ECE from JNTU Kakinada and received MTech also from JNTU Kakinada, Andhra Pradesh. Currently, she is a research scholar of JNTU Kakinada and working as an assistant professor in ECE Department at Chaitanya Bharathi Institute of Technology, Gandipet, Hyderabad, India. Her research interest areas are hyper spectral imaging and classification.

P. V. Naganjaneyulu

P V Naganjaneyulu received PhD in communications from JNTU Kakinada and he is currently professor and principal of Sri Mittapalli College of Engineering, Tummalapalem, Guntur District, A P. He has guided 3 PhD scholars and at present 12 scholars are working with him. He specializes in communication and signal processing. He published more than 50 journal papers and 10 conference papers. His current research area of interest is communications, VLSI signal processing and image processing. He is a member of IETE. Email: [email protected]

K. Babulu

K Babulu has obtained graduation from GITAM Engineering College, Visakhapatnam affiliated to Andhra University in the Department of Electronics and Communication Engineering and master's degree in electronic instrumentation from REC, Warangal, Andhra Pradesh. He did doctoral degree in the field of VLSI & embedded systems from JNT University, Anantapur. He is currently working as professor in Department of ECE, University college of Engineering, JNTU Kakinada. He has published 22 research papers in reputed national and international journals. He shared his research experience on about 35 national and international conferences, workshops, seminars and symposia. He is a member of professional bodies such as ISTE (Life Member), IE(Fellow) and IETE(Fellow). His total teaching and research experience is 15 years. His interested areas are VLSI & embedded systems. Email: [email protected]