作者: Konstantinos G. Derpanis , Richard P. Wildes
DOI: 10.1007/978-3-642-12304-7_29
关键词:
摘要: The detection of motion boundaries has been and remains a long-standing challenge in computer vision. In this paper, the recovery is recast broader scope, as focus placed on more general problem detecting spacetime structure boundaries, where constitute special case. This recasting allows uniform consideration between wider class patterns than previously considered literature, both coherent well additional dynamic patterns. Examples beyond standard that are encompassed by proposed approach include, flicker, transparency various textures (e.g., scintillation). Toward end, novel representation method for these raw image sequence data presented. Central to description oriented distributed manner. Empirical evaluation boundary detector challenging natural imagery suggests its efficacy.