Image Fusion Method Based On Sparse Depth Network

Image fusion is the integration of multiple image information acquired by multiple sensors at the same time into an image containing all important information.and its purpose is to improve the resolution and clarity of the image.Degree to facilitate observation and reprocessing.This paper mainly introduces the image fusion method under sparse representation(SR).Different from the exchange domain and the space domain,the sparse representation is that the signal can be approximately represented by the sparse linear combination of several atomic signals in the over complete dictionary.In addition,the transformation space of sparse representation is sparse domain,and the active measurement in sparse domain is more reliable.The image is segmented by using image block based method through spatial sliding window technology,which can better retain the details of the image and has translation invariance.Convolution sparse representation is the convolution form of sparse representation,which can further improve the image fusion effect by convolution instead of using image blocks to process the image.In this paper,in the process of image fusion,the methods of convolution sparse representation(CSR),morphological component analysis(MCA)and non-subsampled shearlet transform(NSST)are used to improve the clarity of the fused image.The specific research contents of this article are as follows:(1)In order to solve the block effect of sparse representation in the process of image fusion,an image fusion method based on convolutional sparse representation and morphological component analysis is proposed.Based on the advantages of convolution sparse representation,the morphological component analysis model is improved to form a new model of csr-mca,which can realize multi-component and global sparse representation of source image at the same time.The pre-learning csr-mca model is used to obtain the smooth and sparse representation of the detail components of the source image.Then different fusion rules are used to fuse each image component,and the corresponding dictionary is used to stack and reconstruct the fused components to obtain the final fusion image.(2)In order to overcome the pseudo Gibbs phenomenon in multi-scale transformation,an image fusion method based on non-subsampled shearlet transform and convolution sparse representation is proposed.The non down sampling shear wave can decompose the source image,and the low-frequency and high-frequency coefficients obtained contain the featureinformation in the image.Through different fusion rules,the low-frequency and high-frequency coefficients are fused and reconstructed respectively,and the detail information in the source image is effectively preserved.The convolution sparse representation is introduced into the fusion process to avoid the pseudo Gibbs phenomenon and the block effect in the sparse representation method,so as to improve the fusion effect and have better clarity.(3)In order to achieve the integrity of the fusion process,an image fusion auxiliary platform is built to enhance and register the unprocessed image,so as to better read the natural image and medical image;at the same time,the platform includes the traditional fusion algorithm and the algorithm proposed in this paper,which can achieve better image fusion effect.Through the platform,we can more directly show the process steps of image fusion,and effectively carry out human-computer interaction.