Perceived Sharpness-Based Multi-Focus Image Fusion with Uniform Discrete Curvelet Transform

A novel multi-focus image fusion algorithm is proposed, which is called UDCT-PS. This algorithm is developed based on the uniform discrete curvelet transform (UDCT) and the perceived sharpness (PS) measurement. Input images are first decomposed by UDCT which is a true tool of 2-D multi-scale transform. So as to the coefficients of different scales and directions are obtained. And then, an appropriate selection principle of the fused coefficients is built according to the characteristics of multi-focus images and decomposed coefficients. The structural features are contained in the coarsest scale coefficients which are merged to create a fusion principle based on the slope of the local magnitude spectrum (SLMS). Highpass directional subband coefficients represent the details of input images, which are merged by a fusion rule based on the local total variation (LTV). Compared with several state-of-the-art image fusion methods, experimental results demonstrate that the proposed algorithm generates encouraging performance.