Research Of Multi-focus Image Fusion Algoithm Based On TV And GF Under The Framework Of Dual-Scale Decomposition

Multi-focus image fusion is an important branch of image fusion.Due to the limited depth of field of the optical lens,only objects with better focus can obtain a clear image.The fusion processed image not only improves the reliability of the information but also makes full use of the complementary information,thus providing a better perspective for human and machine perception.After studying the related algorithms,this thesis proposes two fusion methods in the framework of dual-scale transformation.The main contents are as follows:First,this thesis briefly introduces several spatial domain and transform domain based fusion methods and rules,and introduces commonly used evaluation indicators and multi-focus visible light imaging characteristics.Second,a fusion method based on total variation(TV)and phase congruency(PC)is proposed.The basic theory of TV and PC and the specific integration steps of the two parts are introduced.Gaussian filter for fast dual-scale decomposition can reduce detection error and image distortion,and greatly reduce the amount of data processing.Since the total variation is robust to noise,this method reduces the sensitivity to sensor noise.Better fusion effect.Through the experimental analysis,the superiority of the method is proved.Thirdly,a fusion method based on Guided Filtering(GF)and Sum Modified Laplacian(SML)is proposed.This part introduces the relevant theoretical knowledge of GF and the superiority of SML in application and the specific fusion process.The method uses a weight construction method based on GF's pixel saliency and spatial background to overcome the problem of brightness and color distortion caused by ignoring spatial consistency.The comparison of the five sets of images with the six algorithms and the measurement of four sets of objective indicators prove that the method is superior in performance.