Abstract: Action Recognition is an active research area in computer vision and among the action recognition tasks Violent action recognition has high importance as it can be used for safety and security enforcement. Design of Intelligent robots and Intelligent Surveillance systems that can automatically detect violences, Violent video indexing systems etc. are the current trends in computer vision and Artificial Intelligence. This paper surveys the different computer-vision based methods available for representing and recognizing violent actions. Core technology beneath every action or activity recognition system is Feature extraction and classification. Feature extraction is the process of fetching those features which are highly discriminative. In case of Violent activity recognition computer vision provides many such feature extraction methods which we can broadly classify as Local and Global features ,Spatial and temporal features. Current trends in Violent action recognition is based on combination of these local, global spatio-temporal features.