Multi-focus Image Fusion Based On DTCWT And Dense SIFT

Vision is an important way for humans to obtain information,and images are also a kind of visual information.In order to meet people's various needs for images,dual-camera phones have become the development focus of various mobile phone manufacturers.According to different needs,dual cameras also provide a variety of combinations,such as black and white cameras,dual color cameras,wide-angle and telephoto cameras and so on.Because of the different functions of the camera,there will be some differences in the pictures taken.Through the multi-source image fusion method,multiple groups of pictures taken in the same scene can be fused to obtain a fusion image with better visual effect,which is an important application of the multi-source image fusion method in daily life.For digital images,pictures taken under fixed conditions are often difficult to reflect all the information of the scene.Limited by the depth of field,only objects at a specific distance from the camera will be focused and clearly captured,while objects located in front of or behind the focusing plane will be defocused or blurred.In order to obtain all the information in a scene,an effective method is to use multi-focus image fusion technology,which means using imaging devices with different focal lengths to obtain multiple sets of focused pictures of objects in various depth ranges,removing redundant information,and finding complementary information.Finally,the photos can be fused into a clear image.This paper first introduces the concept of multifocal image fusion.Then,a simple summary of the steps of the fusion method based on transform domain is made.After that,the theory of the dual-tree complex wavelet transform,Poisson image processing and dense SIFT descriptors used in the paper are briefly introduced.In addition,in view of the feature that the high-frequency sub-band information of the image decomposed by the dual-tree complex wavelet transform is a complex number,a suitable dense SIFT descriptor fusion method is proposed to achieve multi-focus image fusion.Subsequently,the evaluation methods commonly used in multi-focallength image fusion methods are introduced.By comparing with several classical transform domain fusion methods and the fusion method that also uses dense SIFT descriptors,the feasibility of the fusion algorithm proposed in this paper is verified in terms of visual effects and fusion evaluation indicators.Aiming at the problem of large time overhead and memory limitation in the method proposed in this paper,a new low-frequency sub-band processing method is analyzed and introduced,namely the pixel-level visibility method.Later,the fusion rules were redesigned and comparative experiments were conducted.Experimental results show that the improved algorithm has improved visual effects.In terms of fusion indicators,not only have the good indicators of the original algorithm been maintained,but the effects of other indicators have also been improved.At the same time,it solves the problem of limited memory,and it can achieve a better fusion effect for larger gray images and color images.