Research On Multi-focus Image Fusion Algorithm Based On Image Decompositionon

Image fusion plays an important role in the field of image processing,which has received widely attention in other fields.As an important branch of image fusion,multi-focus image fusion overcomes the limitation of depth of field.By fusing images with different depth of the same sensor in the same scene and reconstructing into a clear fusion image,multi-focus image fusion makes all the objects of this scene in focusing.In the environment of energy-limited and real?time application,the nodes are limited by the computation and storage resources.Processing the redundant image information obtained by a single node will generate unnecessary energy consumption,and reduce the working life of the node.Therefore,image fusion technique can usually extract and fuse effective information and remove the distractive information.In order to make the fusion image close to the real scene more natural,the fusion algorithm is usually complicated.However,it will bring more complex computing overhead.Combined with the features of real-time application,two multi-focus image fusion algorithms based on image decomposition are proposed.A novel multi-focus image fusion method based on Split Bregman decomposition has been proposed in this paper.Firstly,the obtained source images are decomposed into structure and texture components by the split bregman decomposition.According to the.feature of these components,more salient features can be extracted based on a new focus evaluation RMD(range-modulus-difference)and obtain the focus map of each source image synthetically.Finally,the initial decision map of all the objects is constructed by comparing the focus map of each source image.With guidance of the initial decision map,the fusion image can be reconstructed.An efficient fusion approach based on FPDEs(fourth-order partial differential equations)for multi-focus images is also proposed.Firstly,decomposing by the fourth-order partial differential equations(FPDEs),the source images are decomposed into structure and texture components.The computational complexity of this process is extremely low.Then,the initial decision map can be construted by fusing these components based on the regional constrast average-range-modified-Laplacian(ARML)method.Finally,the fusion image can be reconstructed by the initial decision map.Comparing with traditional transformed methods like wavelet transform(WT)and other existing spatial methods,the fusion results and computational consumption of these two proposed algorithms is better than others.These algorithms also show better performace both in subjective evaluation and objective evaluation.The experimental results demonstrate that these two proposed fusion approach can achieve better performance in energy-limited environment.