Improving the Coding of Regions of Interest

homergcc.ee.ntu.edu.tw Abstract-This paper considers a video coding system for (through down-sampling) a full-view video at low resolution surveillance applications. It consists of one "base encoder " and an ROI video at the original resolution. Another that encodes a down-sampled, full-view version of the input requirement of the system is that, unlike the conventional video sequence and one "region of interest " (ROI) encoder scalable video coding schemes, each of the compressed that encodes an ROI of the video sequence at the original video bitstreams should be independently decodable. For image resolution. An important requirement of the video this reason, each of the full view video and the ROI videos coding system is that the ROI bitstream and the base bitstream needs one encoder, as shown in Figure 1. The architecture should be independently decodable. We explore the of the surveillance system under consideration is shown in inter-relationship between the full-view video sequence and the Figure 2. The system consists of a base encoder to encode ROI video sequence and apply it to improve the computational the down-sampled, full-view video and an ROI encoder to efficiency of the ROI encoder. The proposed algorithm achieves an average ofl170 % speedup. encode ROI sequence at the original camera resolution.Note that there may be more than one ROI in a video I. INTRODUCTION sequence. For simplicity, only one ROI is considered in this paper. However, the algorithm can be applied t