A Framework of Hyperspectral Image Compression using Neural Networks

Hyperspectral image analysis has gained great attention due to its wide range of applications. Hyperspectral images provide a vast amount of information about underlying objects in an image by using a large range of the electromagnetic spectrum for each pixel. However, since the same image is taken multiple times using distinct electromagnetic bands, the size of such images tend to be significant, which leads to greater processing requirements. The aim of this paper is to present a proposed framework for image compression and to study the possible effects of spatial compression on quality of unmixing results. Image compression allows us to reduce the dimensionality of an image while still preserving most of the original information, which could lead to faster image processing. This paper presents preliminary results of different training techniques used in Artificial Neural Network (ANN) based compression algorithm. Keywords— Image Compression, ANN, HSI Digital Object Identifier (DOI): http://dx.doi.org/10.18687/LACCEI2015.1.1.189 ISBN: 13 978-0-9822896-8-6

[1]  Qian Du,et al.  High Performance Computing for Hyperspectral Remote Sensing , 2011, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

[2]  Cheng-Chang Lu,et al.  A neural network based image compression system , 1992 .

[3]  Qian Du,et al.  Hyperspectral Image Compression Using JPEG2000 and Principal Component Analysis , 2007, IEEE Geoscience and Remote Sensing Letters.

[4]  Linmi Tao,et al.  A robust band compression technique for hyperspectral image classification , 2010, 2010 IEEE International Conference on Intelligent Computing and Intelligent Systems.

[5]  E. Watanabe,et al.  Lossy image compression using a modular structured neural network , 2001, Neural Networks for Signal Processing XI: Proceedings of the 2001 IEEE Signal Processing Society Workshop (IEEE Cat. No.01TH8584).