Abstract:

    A multifocus image fusion algorithm based on directional windows statistics in nonsubsampled contourlet domain is developed. First, source images are decomposed to the domains of nonsubsampled contourlet transform. Then, the image fusion is implemented in subbands with different scales and directions combining with regional statistics. Regional variance and local energy are adopted as the fusion rules in lowpass and highpass subbands, respectively. Finally, the fused image is obtained through inverse transform. Experimental results show that the directional windows in nonsubsampled contourlet can detect the image features more effectively and the fused image has better subjective visual effect. Comparison indicates that the proposed algorithm outperforms the traditional gradient pyramid algorithm and the wavelet based algorithm in terms of the fusion quality index.

Key words: image fusion, nonsubsampled Contourlet transform, multifocus, directional windows