Research On Hierarchical Iterative Image Fusion Algorithm Based On Feature Image Gradient

In recent years,image fusion algorithms have developed rapidly in the fields of industrial production and medicine,especially playing an important role in some multi-scene applications.The source images obtained by different methods have different focal points or focal areas.The information obtained by a single sensor is often not very comprehensive,and the information in different images is complementary.The image fusion algorithm is committed to making the characteristics of different source images as much as possible.The retention of the image in the fused image further improves the spatial perception of the image,facilitates image analysis and processing,and is more suitable for human visual recognition and computer follow-up detection and classification.The research consideration of the algorithm should not only consider the fusion effect,but also consider the universality and real-time.Aiming at the research of image fusion algorithm,based on the ideas of transform domain denoising and spatial domain fusion,an image fusion algorithm based on feature image gradient is proposed.The research results obtained are as follows:1.Use the Joint Sparse Representation(JSR)algorithm to denoise and obtain the common part of the image and the respective feature image of each source image.Observe the public image obtained after denoising and the respective feature image and compare with the source image;compare the features respectively The image is subjected to gradient operations to prepare for the subsequent hierarchical iteration and region division process.This method combines the advantages of transform domain algorithm denoising with the advantages of spatial domain fusion,and achieves good results.2.Use the method of image segmentation and comparison of gradient sum to do basic regional division of the image,use the hierarchical iterative strategy to obtain the final regional decision diagram,and use the regional decision diagram and fusion strategy to obtain the final fusion image.3.The algorithm combines the improved algorithm JSR of sparse representation(SR),and then uses the idea of gradient hierarchical iteration to basically achieve the purpose of dividing clear and fuzzy blocks,and the calculation amount of this process is small.