Automatic focusing method based on structure tensor

The invention discloses an automatic focusing method based on a structure tensor, which relates to the field of computer vision, and is technically characterized in that: for imaging of a digital holographic imaging system, determining the axial position of a recorded object is a key step of image reconstruction. This problem becomes more challenging when multiple discrete portions of a three-dimensional object overlap one another. The invention provides an automatic focusing algorithm based on a structure tensor. According to the method, the focal length of each section of the multi-section object can be extracted without being influenced by whether the sections are overlapped or not. The applicability of the proposed technology is verified by using comprehensive and experimental data of two types of holographic systems. 本发明公开了一种基于结构张量的自动聚焦方法，涉及计算机视觉领域，其技术特征在于：针对数字全息成像系统的成像，确定记录物体的轴向位置是图像重建的关键步骤。当一个三维物体的多个离散部分相互重叠时，这个问题变得更具挑战性。本文提出了一种基于结构张量的自动对焦算法。该方法可以提取多截面物体各截面的焦距，而不受截面是否重叠的影响。我们用两种类型的全息系统的综合和实验数据验证了所提出技术的