3D reconstruction of small sized objects from a sequence of multi-focused images

Abstract 3D reconstructions of small objects are more and more frequently employed in several disciplines such as medicine, archaeology, restoration of cultural heritage, forensics, etc. The capability of performing accurate analyses directly on a three-dimensional surface allows for a significant improvement in the accuracy of the measurements, which are otherwise performed on 2D images acquired through a microscope. In this work we present a new methodology for the 3D reconstruction of small sized objects based on a multi-view passive stereo technique applied on a sequence of macro images. The resolving power of macro lenses makes them ideal for photogrammetric applications, but the very small depth of field is their biggest limit. Our approach solves this issue by using an image fusion algorithm to extend the depth of field of the images used in the photogrammetric process. The paper aims to overcome the problems related to the use of macro lenses in photogrammetry, showing how it is possible to retrieve the camera calibration parameters of the sharp images by using an open source Structure from Motion software. Our approach has been tested on two case studies, on objects with a bounding box diagonal ranging from 13.5 mm to 41 mm. The accuracy analysis, performed on certified gauge blocks, demonstrates that the experimental setup returns a 3D model with an accuracy that can reach the 0.05% of the bounding box diagonal.