针对光学成像系统景深范围有限、景深范围外目标成像模糊的问题, 提出一种区域方差和点锐度相结合的拉普拉斯金字塔变换图像融合方法。该方法首先对多幅源图像做拉普拉斯金字塔变换, 提取源图像在不同频率层的特征信息; 对金字塔顶层图像和其余各层图像, 分别采用区域方差和点锐度作为融合度量, 对其进行融合, 得到融合图像的金字塔各层系数; 最后进行拉普拉斯金字塔逆变换得到融合图像。仿真结果表明, 本文方法得到的融合图像主观视觉效果和客观评价指标均优于其他常用算法, 标准差、互信息和边缘保持度分别提高了0.24%、6.8%和8.4%。该方法有效降低了融合图像的噪声, 获得了更加丰富的边缘信息, 对实现光学成像系统在大视场范围获取高分辨率光学图像具有重要意义。

Depth of field of traditional optical systems is limited, it is hard to achieve a wide field image with high resolution. A novel fusion method of multi-focus image based on Laplacian pyramid transform which evaluated by the metrics of regional variance and Point sharpness (EAV) was proposed. The fusion method firstly performs Laplacian pyramid transform on multiple source images, and feature information at the layers with different frequencies were extracted. Then, the multiple source images were fused by utilizing the metric of regional variance for the top layer, and EAV for the other layers. Finally, the fused image was achieved through the inverse Laplacian pyramid transform. The obtained fused image is with lower noise and more abundant edge information than those achieved by the commonly used methods. After evaluation, the standard deviation, the mutual information and the edge retention are improved by a factor of 0.24%, 6.8% and 8.4%, respectively. It is of great significance for optical systems to achieve images with high-resolution over a wide field of view.