中文摘要：

利用非下采样Contourlet 变换（NSCT）的系数特点, 设计NSCT 域中高低频融合规则, 并结合基于区域分割的边缘检测算法, 针对多聚焦图像提出了一种有效的融合算法. 首先通过NSCT 变换把2 幅待融合图像分解为一个低频系数矩阵和一系列的高频系数矩阵, 对低频系数采用局域方差、局域空间频率、局域改进的拉普拉斯能量和3 个统计特征的加权平均进行融合, 对高频系数基于局部纹理特征进行融合, 经过NSCT 逆变换后得到初步的融合结果;然后根据初步融合图像和待融合图像的残差图识别出清晰区域, 对清晰区域进行边缘检测, 将该边缘信息覆盖到初步融合的图像上, 得到最终的融合图像. 与传统DWT, NSCT 变换和基于视觉特性的NSCT 域图像融合算法进行实验对比的结果表明, 该算法在视觉效果和平均梯度、互信息、空间频率与边缘保持度等多个评价指标上均有优势.

英文摘要：

According to the characteristics of the coefficients of non-subsampled Contourlet transform（NSCT）, new fusion rules for low and high frequency in NSCT domain are developed. Furthermore, a newmulti-focus image fusion algorithm is proposed by combing the edge detection algorithm based on regionsegmentation. Perform NSCT transform on the two source images, the low frequency coefficients and highfrequency coefficients are obtained, the low frequency coefficients are fused by the weighted average of statisticalfeatures of local variance, local spatial frequency and sum of local modified Laplacian energy, whilethe high frequency coefficients are fused by local texture features; then, the preliminary fused image is obtainedby inverse NSCT transform; finally, the clear area is identified according to the residual images of thepreliminary fused image and the source images, edge of clear area is detected, and the final fused image isobtained by covering the edge information to the preliminary fused image. Experiments are conducted tocompare the proposed algorithm with the algorithms of traditional DWT, NSCT transform, and the algorithmof NSCT domain based on the visual characteristics. The experimental results show that the proposed algorithmcan get better visual effect and higher evaluation index values （including the average gradient, spatialfrequency, mutual information and edge preserving degree）.