Multi-focus Image Fusion Based On Image Structure-texture Decomposition

Multi-focus image fusion which is mainly used for the fusion of multiple images that are obtained under the same imaging conditions by the same image acquisition system has an important application in image processing.Due to the focus range,the optical imaging system cannot image all the objects in and out the focus clearly at the same time during the process of image acquisition.Obviously this also increases the difficulty of image processing.But Multi-focus image fusion can solve all these problems and it also can make the processing of image information more efficient and even enhance the reliability and accuracy of the system;meanwhile it also expands the scope of the work.Nowadays multi-focus image fusion technology has been widely used in all areas of our lives.Although image decomposition is the key step of image fusion method based on transformation domain,and the level of image decomposition is closely related to the final fusion result.And traditional decomposition methods do not take into account the nature of the image itself,so the decomposition effect is not good,and it is easy to cause the appearance of mixing phenomenon,besides the decomposition process is complex.In order to solve the above problems,the research contents of this paper are as follows:1)Multi-focus image fusion based on Vese-Osher decomposition modelThis paper presents a multi-focus image fusion method based on VO decomposition model.Firstly,the VO image decomposition model decomposes the source image into two parts: Structure and texture,and the structure and texture of the focus region are extracted by using the improved differential modulus and the focus evaluation function,and then the focus region of the source image is obtained;then the initial decision-making map based on the fusion rules is fused and the final multi-focus image is reconstructed on the basis of morphological processing.2)Multi-focus image fusion based on structure-texture dictionary learning algorithmVO model can separate the structure and texture of the image better,but for the noise of the source images it can't handle well,so this paper combines the dictionary learning algorithm,which can remove the source image noise better,to construct the image cartoon and Texture Dictionary adaptively,and get the high quality decomposed image;and it also combines with the improved Brenner operator focusing evaluation method,which is better for edge information extraction,the characteristics of focus region are found out and the full focus image is reconstructed.3)The two proposed fusion frameworks,together with the traditional Brovey,DWT and DSIFT algorithms and the new CSR and DWT-sr algorithms proposed in recent years,are simulated on the Matlab platform,and the results of the subjective evaluation and objective fusion quality test of the fusion image confirm the superiority of the new fusion framework.The results show that the two methods meet the requirements and provide support for multi-focus image fusion.