Image Clustering and Video Summarization for Efficient 3D Modelling and Reconstruction

Abstract

Although high quality 3D representations of important cultural landmarks can be obtained via sophisticated photogrammetric techniques, their demands in terms of resources and expertise pose limitations on the scale at which such approaches are used. In parallel, the proliferation of multimedia content posted online creates new possibilities in terms of the ways that such rich content can be leveraged, but only after addressing the significant challenges associated with this content, including its massive volume, unstructured nature, and noise. In this chapter, two strategies are proposed for using multimedia content for 3D reconstruction: an image-based approach that employs clustering techniques to eliminate outliers and a video-based approach that extracts key frames via a summarization technique. In both cases, the reduced and outlier-free image data set is used as input to a structure from motion framework for 3D reconstruction. The presented techniques are evaluated on the basis of the reconstruction of two world-class cultural heritage monuments.