Video processing

Video data is used to recognize human activity based on facet model features. The method gives very good results on regular or routine activities like running, walking, jogging, handwaving, etc. This is also applied on Indian classical dance data to recognize different dance styles or schools like Odissi, Bharathanatyam, etc. Sometimes important objects are obscured by undesired objects. A video inpainting algorithm is developed to solve this problem and predict the desired part of the scene behind the undesired objects. This video inpainting method makes use of higher order SVD for predicting the target patch, and produces much more reliable output.