Research On Multi-focus Image Fusion Based On Decision Map And Sparse Representation

The goal of multi-focus image fusion is to extract clear regions in each image of multi-focus image,and merge them into a fused image with all scenes.How to locate in-focus regions of source images accurately and effectively is the key to fuse multi-focus images.However,it is not easy for most of traditional multi-focus image fusion methods,considering that the content and structure of multi-focus images are always complex.Aiming at the shortcomings of existing multi-focus image fusion methods,this paper proposes a new multi-focus image fusion framework-the multi-focus image fusion algorithm based on decision map and sparse representation(DMSR).In this algorithm,the clarity level of the low scale image of source images is measured with sparse representation and an effective clarity level measurement,because evaluating the clarity level of source images directly increases the complexity of algorithm and the difficulty of feature extraction.Each pixel of source images is classified into three possible categories:focused,defocused or uncertain.And then,uncertain pixels are further classified by using the spatial frequency method to obtain decision map.Finally,extracted in-focus regions can be merged into the fused result based on the final decision map.Unfortunately,there may be some pixels in transitional regions which are hard to be classified accurately would reduce the image fusion quality.The image fusion method based on sparse representation is applied to process transitional regions.After combining the fused result of decision map and the fused result of transitional regions,the final fused image can be generated.The theoretical analysis and experimental results show that the proposed fusion frame is reasonable and effective.The simulation results on three multi-focus image data sets show that the proposed method is more efficient than the traditional methods.