基于结构相似指数的聚焦形貌恢复

Shape from focus based on structural similarity

针对目前大部分聚焦形貌恢复算法对边缘部位恢复效果欠佳的问题,从全局入手,考虑空间连续性约束,提出一种新的基于结构相似指数的聚焦形貌恢复算法。根据已有的部分聚焦的图像序列,得到一幅全聚焦图像,然后优化一个基于结构相似指数、全聚焦图像和空间连续性的损失函数,最终重构表面形貌。实验表明,提出的算法能够较好地反映物体的表面形貌信息,对于边缘区域的重构效果较好。

Since most Shape From Focus （SFF） methods have poor effect on edge areas. Taking a whole image and spatial consistency into account, a new shape from focus algorithm was presented based on Structural SIMilarity index （ SSIM）. Different from the existing SFF methods, a full focused image was got from existing partially focused image sequences with varying focus planes, and the surface was reconstructed by optimizing a cost function based on structural similarity index, full focused image and spatial consistency. The experimental results show that the presented algorithm can reflect the surface information especially edge areas more effectively.