Research Of Multi-focus Image Fusion Algorithm

The cross-development of multi-disciplinary network radiation leads to a closer relationship between disciplines,which makes the methods of different disciplines can be used for reference.For example,the images obtained by physical imaging equipment must be converted into digital images for computer use in order to facilitate human processing,and the data in digital images can be displayed concretely in the computer by means of mathematical methods,which enables people to intuitively understand the contents and attributes of different images.Images acquired in the same scene under different conditions always have a certain degree of information loss.As an effective solution,image fusion can synthesize multiple images with less information into a single image for subsequent human recognition and computer processing.According to image acquisition methods and application scenarios,image fusion can be divided into several application fields,such as remote sensing,infrared,visible light and medical images.For specific fields,images obtained by different imaging devices have different properties,such as infrared images can not be affected by weather and obstacles,visible light images are based on the principle of light reflection,so they have rich spatial features and spectral information,while remote sensing images generally have high spatial resolution due to their special functions.How to improve the effectiveness of image fusion in the same scene under different conditions has become one of the key issues in the research of image fusion task.In this paper,aiming at the problems of blocking effect,artifact,distortion and fusion efficiency encountered in the process of multi-focus image fusion,Gaussian filter and guided filter are used to improve the above problems.The main work is as follows:Firstly,the existing image fusion methods are roughly divided into two parts according to the different processing methods: the fusion method based on transform domain and the fusion method based on spatial domain,and then the divided two parts are further refined according to the size and adaptability of the processed objects.In this process,the rapid development brought about by the introduction of two technologies is emphatically introduced: the introduction of multi-scale geometric analysis research and the introduction of deep learning methods.At the same time,with the development of fusion technology,the standard for evaluating the quality of a fused image has become a hot spot for many scholars.According to some specific features of the image,the related evaluation criteria are constantly improved and updated,thus reflecting the performance of the fusion algorithm more comprehensively and accurately.Secondly,aiming at the common problems of uneven exposure and incomplete focusing in photography,image fusion technology gives corresponding solutions.In this paper,the related technologies are improved aiming at the bad situation in the process of multi-focus image fusion.Firstly,the attributes of the focused image are analyzed.According to the basic fact that the focused part of the image is often the high-frequency component in the image,a Gaussian low-pass filter function is proposed to filter out the low-frequency component of the image,and then the difference between the source image and the low-frequency component is used to take its absolute value to reflect the focusing information of the image.The edge-preserving characteristic of guided filtering is used to achieve the effect of smoothing the boundary between focused area and defocused area,so that the synthesized partial transition in the fused image is more natural.Thirdly,the work done in this paper is summarized from the content and method,and new methods are put forward to improve the imperfectness of objective evaluation standard data in the fusion results.Finally,the research direction of future fusion algorithm is prospected from the expansion of application field and the real-time performance of evaluation algorithm.