| With the availability of multi-sensor,multi-time,multi-resolution and multi-spectral image data from Earth observation satellites,image fusion has become an important tool in remote sensing image evaluation.Due to the limitations of satellite storage space,satelliteto-ground data transmission speed,and energy entering the sensor within the same time interval,most satellites cannot directly collect high spatial resolution multispectral images to meet the requirements of high spatial and high spectral resolution.Usually,the image obtained by the multi-spectral sensor has high spectral resolution,and the image obtained by the panchromatic sensor has high spatial resolution.Use appropriate fusion algorithms to combine the images obtained by the two sensors to produce the best features of both the image is of great significance for improving the spatial resolution of multi-spectral images,improving the utilization and recognizability of images,improving the visual effects of images,and enhancing image features.Therefore,this paper mainly focuses on the fusion method of panchromatic image and multispectral image.The main research contents and results of this article are as follows:(1)Two-dimensional local mean decomposition is another multi-scale analysis tool after Wavelet,Contourlet,and BEMD,this method is suitable for nonlinear and nonstationary signal analysis,and can realize multi-scale decomposition of signals.As a brandnew multi-scale analysis method,it provides a new perspective and new ideas for the fusion of two-dimensional image signals.Therefore,this paper chooses the newly emerging spatial multi-scale decomposition technology BLMD to study its fusion application in panchromatic images and multispectral images.Using images of different regions in GF-1and ZY-3 as data sources,fusion experiments were carried out.The fusion results of BLMD method and traditional fusion methods were compared and analyzed,and the method was evaluated for classification.The qualitative evaluation results show that the BLMD method has more consistent spectral information with the reference image,but the ability to retain the original image spatial information needs to be improved,and there is still some gap between the fusion result and the reference image;the quantitative evaluation result shows:the spectrum of the BLMD fusion method Both retention and spatial information injection are better than traditional fusion methods,and the classification accuracy is higher.The feasibility of this fusion method further expands the image fusion system based on multiresolution analysis.(2)Aiming at the problem of insufficient spatial information injection in the twodimensional local mean decomposition fusion method,considering that the imaging mechanism of PAN image and MS image is the same,although the reflectivity of ground objects to electromagnetic waves of different wavelengths is different,the spectral information in the image is different,but The spatial information is the same.In this paper,the Least Squares Theory(LS)is introduced on the basis of two-dimensional local decomposition.It is believed that the high-frequency components obtained after BLMD decomposition of multi-spectral and panchromatic images are correct.Observations of the same target with different accuracy can be optimally estimated by the least square method,so as to obtain high-frequency components with higher reliability and richer details,that is,spatial information.In this paper,four experimental areas containing buildings,roads,water bodies,and vegetation in ZY-3 and GF-2 images are selected for BLMD fusion experiments based on the principle of least squares.The results show that the BLMD fusion method based on optimal estimation can better integrate The spatial information of panchromatic and multispectral images solves the problem of insufficient spatial information injection in the BLMD method,so as to obtain more ideal high spatial resolution multispectral images. |
PDF Full Text Request