Abstract: A multi-focus image fusion method based on information bottleneck clustering is proposed.The method attains the joint clustering denotation through the clustering analysis of source images using information bottleneck algorithm.After the source images are decomposed using non-sampled contourlet transform，the high/low frequency coefficients are fused by the guide of the joint clustering denotation.The fused image is reconstructed in using non-sampled contourlet transform.The experimental results show the proposed method possesses favorable objective evaluation performance and subjective vision effect.

Key words: image fusion, information bottleneck method, non-sampled contourlet transform

摘要： 提出一种基于信息瓶颈聚类的多聚焦图像融合方法。该方法采用信息瓶颈算法对源图像进行聚类分析，获得联合的聚类表示；由非下采样Contourlet变换对源图像进行多分辨率分解，通过联合聚类表示指导各频域系数融合；采用非下采样Contourlet逆变换重构获得融合图像。实验结果表明，该方法具有良好的客观评价性能和主观视觉效果。

关键词: 图像融合, 信息瓶颈算法, 非下采样Contourlet变换