Research in Appearance Description for Machine Vision.
暂无分享,去创建一个
Abstract : The key barrier to application of machine vision in unconstrained environments is the complexity of image formation in the world and the resultant difficulty of characterizing it concisely. If we could create a general yet concise description of image formation, we would have a vocabulary for discussing the complexity of specific scenes and the assumptions of specific machine vision approaches. In this research, the investigators are attempting to develop such a 'vocabulary' consisting of a mathematical formalism for describing scenes, and examples of programs that utilize this formalism and data that is described using it. The data collection is not only for the purpose of this research contract, but also as a way of archiving and broadcasting high-quality image data from the Calibrated Imaging Laboratory at CMU, that may be useful for other researchers in image understanding.