Scanning electron microscopes (SEM) allow a detailed surface analysis of a wide variety of specimen. However, SEM image data does not provide depth information about a captured scene. This limitation can be overcome by recovering the hidden third dimension of the acquired SEM micrographs, for instance to fully characterize a particle agglomerate's morphology. In this paper, we present a method that allows the three-dimensional (3D) reconstruction of investigated particle agglomerates using an uncalibrated stereo vision approach that is applied to multiple stereo-pair images. The reconstruction scheme starts with a feature detection and subsequent matching in each pair of stereo images. Based on these correspondences, a robust estimate of the epipolar geometry is determined. A following rectification allows a reduction of the dense correspondence problem to a one-dimensional search along conjugate epipolar lines. So the disparity maps can be obtained using a dense stereo matching algorithm. To remove outliers while preserving edges and individual structures, a disparity refinement is executed using suitable image filtering techniques. The investigated specimen's qualitative depth's information can be directly calculated from the determined disparity maps. In a final step the resulting point clouds are registered. State-of-the-art algorithms for 3D reconstruction of SEM micrographs mainly focus on structures whose image pairs contain hardly or even none-occluded areas. The acquisition of multiple stereo-pair images from different perspectives makes it possible to combine the obtained point clouds in order to overcome occurring occlusions. The presented approach thereby enables the 3D illustration of the investigated particle agglomerates. © 2018 SPIE.