Fusion Method Of Multispectral Image Based On Non-Separable Wavelet Of Quincunx Sampling

With the rapid development of the remote sensing technology, multispectral image fusion technology has been concerned and applied widely. Its theories and methods have become an important research realm of the subjects such as intelligent information processing, and remote sensing application, etc. The fusion of multispectral image integrates the images which have higher spectral quality but lower spatial resolution and the panchromatic images with higher spatial resolution. It creates a new image that has better spectral information and higher spatial resolution.First of all, this thesis briefly describes the definition, steps, levels and applications of the multispectral image fusion, as well as the current domestic and foreign research on it. Secondly it systematically summarizes the current multispectral image fusion methods and evaluation standards. Regarding the deficiencies of two-dimensional separable wavelets in sampling only prominent vertical and horizontal directions, non-separable wavelet of quincunx sampling’s characteristics and advantages is explained, and then two methods of multispectral image fusion which based on non-separable wavelet of quincunx sampling are proposed:One is based on HSV transform and Red-Black wavelet transform. A method of image fusion based on HSV transform and Red-Black non-separable wavelet transform is presented and its application in the fusion of multispectral image and panchromatic image. Firstly, the multispectral image is transformed by HSV transform, multi-resolution Red-Black wavelet decomposition of the value component of multispectral image and panchromatic image is performed. Then different fusion algorithms are used to the high-frequency parts and the low-frequency part. Finally, the fusion result image was obtained by Red-Black reconstruction and inverse HSV transform. The fusion performance of the fusion method is evaluated using objective index. The experiment results show that this method have good visual effective in the preservation of spectral quality and spatial resolution information. This fusion method can preserve good spectral information and get more information from the source images. Moreover it can also preserve higher spatial resolution information. The other is based on IHS transform and lifting quincunx wavelet transform. A novel fusion method of multispectral image and high spatial resolution panchromatic image based on IHS transform and lifting quincunx wavelet transform which using Neville filter is presented. Firstly, the multispectral image is transformed by IHS transform, multi-resolution lifting quincunx wavelet decomposition of the value component of multispectral image and panchromatic image after performed histogram matching is performed. Then different fusion algorithms are used to the high-frequency parts and the low-frequency part. Finally, the fusion result image was obtained by lifting quincunx reconstruction and inverse IHS transform. The fusion performance of the fusion method is evaluated using objective index. The experiment results show that this method can preserve good spectral information from the value component of multispectral image and fusion higher spatial resolution information from panchromatic image. Its fusion algorithms and the selection of decomposition level are simple and effective for multispectral image fusion.