Multifocus image fusion based on nonsubsampled Bandelet transform

A novel fusion algorithm based on nonsubsampled Bandelet transform is proposed for the multifocus images of same scene. First, the average image is calculated from the source images, after that the best geometric flow is computed using the wavelet coefficients of the image and a Bandelet basis is constructed. Secondly, the source images are represented using the Bandelet basis, and then the obtained coefficients are fused respectively according to their physical significance. Finally, the fused coefficients are applied on the inverse nonsubsampled Bandelet transform to obtain the fused image. The experiment results indicate that, compared with the fusion methods based on Wavelet transform and nonsubsampled Wavelet transform, respectively, the proposed method can provide better fusion quality in terms of both visual sense and quantified measure.