Real-Time Object Segmentation in Smart Camera for Remote Surveillance Scenario

Smart cameras are expected to be important components in video surveillance applications. In any remote surveillance scenario, smart cameras have to take intelligent decisions to select frames of interest and regions of interest to minimize communication and processing overhead. This paper proposes a clustering based scheme for filtering frames of interest and regions of interest using object segmentation to reduce the volume of data needed for further processing and communication. The algorithm is structured for facilitating easy implementation in embedded environment. This scheme enables the camera to communicate only segmented regions in frames of interest to a server and at the same time alleviate the need for computationally costly compression and decompression operations at the source and destination for subsequent tasks like tracking etc. Experimental results have established effectiveness of our scheme.