We propose and investigate a new digital method for the reduction of twin-image noise from digital Fresnel holograms. For the case of in-line Fresnel holography the unwanted twin is present as a highly corruptive noise when the object image is numerically reconstructed. We propose to firstly reconstruct the unwanted twin-image when it is in-focus and in this plane we calculate a segmentation mask that borders this in focus image. The twin-image is then segmented and removed by simple spatial filtering. The resulting digital wavefield is the inverse propagated to the desired object image plane. The image is free of the twin-image resulting in improved quality reconstructions. We demonstrate the segmentation and removal of the unwanted twin-image from in-line digital holograms containing real-world macroscopic objects. We offer suggestions for its rapid computational implementation.