ABSTRACT

ABSTRACT:   A novel method based on the defocus map estimation and the Non-Subsampled Contourlet Transform (NSCT) is proposed in this paper. First, the degree of defocus of source images is estimated. Second, source images are divided into the left focused area, right focused area, and uncertain area, according to the defocus map estimation of source images. At the same time, high-frequency coefficients and low-frequency coefficients of source images are obtained after the NSCT is employed for source images. For uncertain area, Sum of Modified Laplace (SML) and Spatial Frequency (SF) are employed to fuse low-frequency coefficients and high-frequency coefficients, respectively. For the left focused area and the right focused area, the corresponding focused areas of source image are selected directly. Finally, through the combination of the fused uncertain area, the left focused area and the right focused area of the fused image are achieved. Compared with the traditional methods, experiment results show that the fusion effect is improved obviously in vision and parameters.