Research On Heterogeneous Remote Sensing Image Fusion Algorithms Based On IHS＿NSST And Adaptive PCNN

Image fusion technology can achieve the complementary advantages of multi-source images when the original image cannot effectively display information,so that the fused image can better to present useful information,so it has received widespread attention.So far,image fusion algorithms are many and varied.Among them,the fusion method based on multi-scale transform fully considers the difference between the essential attributes of images,and multi-scale and multi-directional decomposition of the images to be fused,which can achieve better results than traditional pixel-level fusion,and is a hot in the field of heterogeneous remote sensing image fusion.Synthetic aperture radar and multi-spectral images have completely different expressions in imaging mechanism and characteristics,and their respective features are distinct and highly complementary.Therefore,it is very suitable for multi-scale fusion algorithm,but currently common multi-scale fusion algorithms still have problems such as low precision and low efficiency.Based on the existing fusion algorithms,this thesis deeply studies the SAR and multi-spectral image fusion based on multi-scale transform.The main research contents and innovations are summarized as follows:(1)In the whole framework of image fusion technology,the fusion algorithms based on IHS transform,wavelet transform,non-subsampled contourlet transform pulse coupled ceural cetwork and non-subsampled shearlet transform are studied respectively,and their differences in fusion algorithms and fusion applications are compared.Then from the subjective and objective point of view,the image fusion quality evaluation system is analyzed in detail.Finally,the advantages and disadvantages of the four fusion methods are illustrated by experiments,and the NSST has better performance in improving the spatial resolution and spectral fidelity of the fused image.(2)The complementary characteristics of IHS,NSST and PCNN are analyzed,and the fusion algorithm combining the three is studied.IHS can separate the intensity component in multi-spectral images,but the spatial details are insufficiently characterized.NSST has a more detailed description of the high frequency subband,but will lose image intensity information.The introduction of the PCNN model that conforms to human visual characteristics can effectively improve this problem.In view of this,the three are combined and applied to SAR and multi-spectral image fusion.(3)In order to solve the problem that the traditional pulse coupled neural network directly uses the pixel value as the neuron input,it is easy to cause the image to be missing in the details,so a high frequency subband stratification fusion rule based on adaptive PCNN is proposed.For high frequency part,high frequency inner subband fusion of the sum of the energy of modified Laplacian as the PCNN input and the modified spatial frequency is used as the link strength of PCNN.Then,the outer subband of the high frequency is combined by the maximal absolute value.Experimental results show that compared with the traditional PCNN fusion,the algorithm can preserve the edge detail information of the source image better in subjective effect.On the objective index,the image obtained by the algorithm is in E,SD,AG,RMSE and CC.The performance of the quality evaluation parameters is superior,and the noise can be suppressed to a certain extent,which verifies the validity and reliability of the proposed algorithm.