Image Fusion Algorithm Combining Structure Transfer And Saliency Detection

It is generally believed that the purpose of image fusion is to fuse as much information as possible about the contours,textures,and gray distributions in original images into the fused image.Now that most existing fusion methods equally process the information from different source images by some feature extraction during the fusion process,however,in most cases it is not suitable for source images with complementary characteristics from different sensors,such as infrared and visible light images.Infrared images usually have a lower spatial resolution than visible light images.Appearance features such as textures in visible light images are often lost in corresponding infrared images because textures do not substantially affect the heat emitted by objects.Therefore,for infrared and visible images,the final fusion image needs to retain both the thermal radiation and the detailed texture information of the source image,which is also the focus of this study.The main research contents and innovations of this dissertation are summarized as follows:1.Through image fusion experiments,the advantages and disadvantages of structure transferring fusion and image-based saliency fusion algorithms are analyzed.The advantage of the structure transfer algorithm is in that it processes the information of infrared and visible light images in different ways,while retaining the advantages of the source images.The structure transfer model in this method is more sensitive to the structure information of the image and can extract the structure information of the image.However,some details of the source image are not kept well.The fusion algorithm based on saliency detection can effectively extract the saliency information of the complementary source image,so that the final fused image has better visual quality,and at the same time can ensure the integrity of the image information.2.Based on the respective advantages of the two algorithms above mentioned,an image fusion algorithm combining the structure transferring and the saliency detection is proposed.This algorithm not only has good structure retention characteristics,but also can retain the saliency details of the source image.At the same time,the feasibility and effectiveness of the algorithm are proved through multiple sets of experiments.3.The effects of four key parameters on the performance of image fusion are analyzed through image fusion experiments,and the suitable parameters are determined.Regularization parameters and window radii are parameters of the structure transfer model,and they affect the similarity between the fused image and the infrared,and visible images.The window sizes of the mean and median filters are different,which can affect the performance of the saliency detection method.4.In order to compare the performance of image fusion algorithms,this paper designs and implements an image fusion system in MATLAB environment.This system integrates the existing classic image fusion algorithms and the algorithms proposed in this paper.By this system different fusion methods can be applied for different types of images or different needs.Also,this system includes the normally used evaluation criterions such as mean,standard deviation,information entropy,and average gradient,which can be applied to compare the performances of different image fusion algorithms in an objective and quantitative manner.