Dense 3-D Structure from Image Sequences Using Probabilistic Depth Carving

摘要: We describe an algorithm to determine dense 3-D structure in a static scene from image sequence captured by moving camera. Metric camera motions are first determined using recursive motion based on tracked feature points. Dense depth information for subset of key frames is then obtained novel probabilistic carvingalgorithm - analogous space carving which probabilities locally about the combined space. An important component this process that opacity and occlusion relationships modelled explicitly, enabling consistent combination probabilities. Results experiments real illustrate effectiveness approach.