基于非下采样Contourlet变换多聚焦图像融合

Fusion of the Multifocus Images Based on the Nonsubsampled Contourlet Transform

针对同一场景多聚焦图像的融合问题,本文提出了一种基于非下采样Contourlet变换(NSCT)多聚焦图像融合算法。首先,采用NSCT对源图像进行多尺度、多方向分解,得到低频子带系数和各带通方向子带系数;其后,针对低频子带系数的选择,提出了一种基于方向向量模和加权平均相结合的融合规则;然后,针对带通方向子带系数的选择,提出了一种基于改进的方向对比度和局部区域能量相结合的融合规则;最后,经NSCT逆变换得到融合图像。实验结果表明,该算法能够有效地保留源图像的有用信息,避免噪声、虚影等效应,是一种有效可行的图像融合算法。

For the fusion problem of the multifocus images with the same scene, a novel multifocus image fusion algo rithm based on the nonsubsampled contourlet transform （NSCT） is proposed. Firstly, source images are decomposed at dif- ferent scales and directions by NSCT, thus the low frequency subband coefficients and various bandpass directional subband coefficients are obtained. Secondly, for the low frequency subband coefficients, we present a fusion rule based on the direc tional vector normal combined with the weighted average; while for the bandpass directional subband coefficients, we pres ent a fusion rule based on the improved directional contrast combined with the local area energy. Finally, the fusion image is obtained through the inverse NSCT. The experimental results illustrate that the proposed algorithm is effective for retaining the original images~ information and avoiding artifacts.