Research On Hyperspectral Image Fusion Based On Wavelet Theory

Hyperspectral remote sensing is looked as a revolution of remote sensing technology. Especialy the invention of imaging spectrometer, hyperspectral image data with higher spectral resolution have become available. Hyperspectral images are gotten from our interesting object with imaging spectrometer by using lots of very narrow Hertzian waves, which make us detect the object by hyperspectral images, but can not be detected by multi-bands images. In order to get the full and whole explain information from hypercpectral image, we must gather the characteristics of object into a single image. This is the hot topic that I want to research—hyperspectral image fusion. Because hyperspectral images have unique characteristics such as enormous data volumes, so the image fusion technology is different from multi-bands image. For getting the object's characteristics that we need, image fusion is a key technique of hyperspectral data processing. These research works, especially the image fusion, have been hot topics in the field of remote sensing. Wherefore, based on wavelet transform, some important aspects of hyperspectral image fusion is studied in this thesis.This thesis mainly talks about three parts:In order to extract features from hyperspectral image data, the principle of hyperspectral images and characteristics are anatomized. In this thesis, use the Airborne Visible and Infrared Imaging Spectrometer (AVIRIS)hyperspectral sensor data. The AVIRIS sensor, which simultaneously collects information in 224 spectral bands that range from 0.4 to 2.5 urn in approximately 10 nm increments, produces 224 images, each representing a single spectral band. Because hyperspectral images have a large number of bands, and has intensity correlation and lots of redundance. Therefore, two effective methods of dividing all bands are proposed according to correlation of neighboring bands and spectral characteristics. At the same time, for different application, the method of selecting bands is proposed.Image fusion is to wipe off the redundant information and combine complement information from hyperspectral images of the same scene, so as to form a single image more suitable for human and machine perception or further image processing tasks. The outline of our research is to put forward a solution ofhyperspectal image fusion that is based on analyzing current induction of the methods of image fusion. At first, all kinds of image fusion methods are summarized and compared. We analyze the applicability of the different image fusion methods for hyperspectral images. Finally, wavelet image fusion is selected in this thesis, and standard variance is used to evaluate the result of fusion.In this thesis, a new wavelet-based hyperspectral image fusion technique with different variance weights is proposed. This new method consists of two key techniques: local feature extraction and weights determination. In order to testify the effectiveness of the proposed method, computer simulations are conducted on AVIRIS data. The fusion algorithm consists of three stages. First, wavelet is used to perform a multiresolution decomposition of each spectral image. Next, the coefficients from each image are combined using a variation-based weighting. The weighting of each coefficient, from a given spectral band image, is determined by the value of each band's variance. The spectral image with the higher value, will receive the larger weight. One spectral image has a higher variance value, which means the image has abundant spacial information and spectral information. Finally, the fused coefficients are used for reconstruction to obtain the fused image. By Comparing with the traditional wavelet fusion techniqueb without weighting, the new technique is show to provide more excellent result. Furthermore, on the basis of two kinds of dividing image spectral band and selecting special bands, the fusion technique with weights is used, we can get more excellent fusion results than that of previous fusion methods.The last chapter summarizes the whole thesis and looks forward the development foreground of hyperspectral image fusion.