Title :
Automatic region of interest extraction and tracking in coded video sequence
Author_Institution :
Sch. of Comput. Sci. & Technol., Southwest Univ. for Nat. (SWUN), Chengdu, China
Abstract :
Region of Interest (ROI) plays an important role in many video-based applications such as video monitoring, object tracking, object-based video coding, etc. However, it is still hard to define, extract and track the ROI in a video sequence. A combined scheme is proposed to automatically determine, extract and track the ROI in a coded video sequence. Both the visual attention model in human vision system (HVS) and the temporal motion cues in a video sequence are incorporated to generate the saliency maps and to determine the ROI. The identified ROI is tracked frame to frame in the compressed domain, using only the motion vectors. This idea can be applied to ROI based video coding, adaptation and content delivery, etc.
Keywords :
data compression; feature extraction; image motion analysis; image sequences; object tracking; video coding; HVS; ROI identification; coded video sequence; compressed domain; human vision system; motion vectors; region-of-interest extraction; region-of-interest tracking; saliency maps; temporal motion cue; video-based applications; visual attention model; Cameras; Feature extraction; Image color analysis; Tracking; Video coding; Video sequences; Visualization; Automatic Region of Interest (AROI); Feature Map; Object Tracking; Visual Attention Model;
Conference_Titel :
Control Science and Systems Engineering (CCSSE), 2014 IEEE International Conference on
Print_ISBN :
978-1-4799-6396-6
DOI :
10.1109/CCSSE.2014.7224532
