After a classifier is trained using a machine learning algorithm and put to use in a real world system, it often faces noise which did not appear in the training data. Particularl...
In this paper a real-time anomaly detection system for video streams is proposed. Spatio-temporal features are exploited to capture scene dynamic statistics together with appearan...
In real world, a scene is composed by many characteristics. Intrinsic images represent these characteristics by two components, reflectance (the albedo of each point) and shading (...
—In real world machine vision problems, numerous issues such as variable scene illumination make edge and object detection difficult. There exists no universal edge detection met...
— To achieve seamless integration of the real scene with virtual objects, we need to obtain accurate illumination condition of the real scene. Using HDRI (High Dynamic Range Imag...