Digital holography is an imaging process able to recreate three-dimensional representations of objects from recording pattern interference among distinct waves. The in-line configuration setup is a variant considered the simplest physical implementation, providing a feasible manner for acquisition and the same time higher resolution for free-living microscopy imaging using a single illumination system. However, the well-known twin-image problem is bounded to the technique, since there is no separation among reference and objects beams in this configuration. As a result, computational numerical diffraction routines present the twin-image effect intrinsically, imposing several difficulties in terms of post-processing requirements. In this context, this paper aims to present a numerical approach able to provide consistent suppression of twin-image problem for in-line holography, during the numerical diffraction procedure for phase retrieval, combining image subtraction and edge detection techniques. The proposed solution was implemented in Python language, and metrics defined to assess it were both qualitative and quantitative, based on edge detection and some image comparison metrics. The obtained results of the proposed approach present a significant reduction in the twin-image artifacts in the reconstructions of both experimental and simulated holograms, considering a spherical reference wave, regardless of the shapes and sizes of objects.

数字全息术是一种成像过程，能够通过记录不同波之间的图案干涉来重建对象的三维表示。在线配置设置是一种被认为是最简单的物理实现方式的变体，它提供了一种可行的采集方式，同时还为使用单个照明系统的自由显微镜成像提供了更高的分辨率。然而，众所周知的双图像问题受到该技术的限制，因为在这种配置下参考光束和物镜光束之间没有分隔。结果，计算数值衍射程序本质上呈现了双像效应，在后处理要求方面施加了一些困难。在这种情况下，本文旨在提出一种数值方法，该方法能够在相变的数值衍射过程中，结合图像减法和边缘检测技术，为在线全息照相提供一致的双图像问题抑制。所提出的解决方案以Python语言实现，基于边缘检测和一些图像比较指标，用于评估该指标的指标既定性又定量。在考虑球形参考波的情况下，无论物体的形状和大小如何，所提出方法的结果在实验和模拟全息图的重建中都显着减少了双像伪像。所提出的解决方案以Python语言实现，基于边缘检测和一些图像比较指标，用于评估该指标的指标既定性又定量。在考虑球形参考波的情况下，无论物体的形状和大小如何，所提出方法的结果在实验和模拟全息图的重建中都显着减少了双像伪像。所提出的解决方案以Python语言实现，基于边缘检测和一些图像比较指标，用于评估该指标的指标既定性又定量。在考虑球形参考波的情况下，无论物体的形状和大小如何，所提出方法的结果在实验和模拟全息图的重建中都显着减少了双像伪像。