基于多尺度局域加权方差的多聚焦融合方法

Multi-focus Fusion Method Based on Multi-Scale Local Weighted Variance

针对在用短波红外显微镜观测目标时,目标的尺寸往往会受到镜头景深限制的问题,提出一种多聚焦融合的方法,可有效扩展显微镜成像的景深。通过改变物距获得大量不同聚焦平面的图像,利用多尺度和局域加权方差来量化区域清晰度,获得焦平面掩膜,之后采用形态学方法优化掩膜边界,最终加权融合焦平面区域得到细节完备的目标整体图。实验中利用景深为10μm左右的显微镜镜头得到了纵向尺寸为几百微米目标的高质量图像。实验结果表明所提方法相较其他方法在对焦平面边界,细节的保留方面有一定优势。

When observing a target with a short-wave infrared microscope,the vertical size of the target is often limited by the depth of field of the lens.Aim at this problem,a multi-focus fusion method that can effectively expand the depth of field of microscope images is proposed.A large number of images with different focal planes is obtained by changing the object distance.Multiscale and local weighted variance is used to quantify the area definition.Then the focal plane mask is obtained.The morphological methods is used to optimize the mask boundary.Finally,focal plane regions are used to obtain the fusion image with complete details by weighted fusion.In the experiment,a microscope lens with a depth of field of about ten microns is used to obtain a high-quality image of a target with a vertical size of several hundred microns.The experimental results show that compared with other methods,the proposed method has certain advantages in terms of keeping the boundary of the focus plane and the details.