Research On Medical Image Fusion Quality Improvement Algorithm Based On Discriminative Low-rank Sparse Dictionary Learning

Medical image fusion plays an important role in image-guided medical diagnostic and treatment.In practice,the images often corrupted by noise during acquisition or transmission.However,most existing image fusion methods are exploited based on the assumption that the source images are noise free.For noisy image fusion,an intuitive method is to denoise the source images,and then fuse them.During this process,the edges and details of source images will be smoothed,that will degrade the visual quality of the final fused result.In this paper,a series of studies have been carried out to improve the quality improvement of medical image fusion,and the results are as follows:(1)most of the methods currently assume that the source images are noise-free and usually do not exist in practice.The performance of traditional fusion method is significantly reduced when the image is disturbed by noise.In this paper,a medical image fusion,denoising and enhancement method based on low rank sparse discriminant dictionary learning is proposed.Specifically,in order to improve the recognition ability of learning dictionaries,we introduce a low order sparse regular term into the dictionary learning model.In addition,in the image decomposition model,we use the weighted kernel norm and sparse control constraints to remove the noise and keep the texture details in the sparse component.Finally,the fusion results are constructed by combining the low rank of the source image and the sparse component.The experimental results show that the method is superior to the existing method in visual and quantitative evaluation.(2)In the study of medical image fusion,the source image should not introduce any artifact or noise,otherwise the result of fusion is often unsatisfactory.In order to preserve the fine scale information of the source image,a multi-component discriminative dictionary learning algorithm and an external block priori guided medical image fusion algorithm are proposed.In this process,a new discriminative dictionary learning model is established.In order to improve their discriminating ability,considering the linear correlation between low-rank components,we add class consistency constraints into coding coefficients to further improve the discriminative ability of dictionaries.In image decomposition,the image decomposition model is established by using the prior and internal self similarity of the external block of the source image.The experimental results show that this method is superior to some advanced methods in objective evaluation and subjective vision.(3)In this paper,the method of medical image fusion,denoising and enhancement based on low rank sparse discriminant dictionary learning is well preserved,but the fine scale information of the image is smoothed.For this reason,a medical image fusion algorithm with multi component discriminant dictionary learning and external block prior guidance is proposed.In this process,not only the rough scale information of the image is preserved,but the fine scale of the image is protected to a great extent.