1. A novel error resilient scheme based on ROI for conversational video transmission
- Author
-
Zhang Can, Chen Deyuan, and He Zhitao
- Subjects
Weight value ,Computer science ,business.industry ,020208 electrical & electronic engineering ,ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION ,02 engineering and technology ,Video transmission ,Motion vector ,Coding block ,Packet loss ,0202 electrical engineering, electronic engineering, information engineering ,020201 artificial intelligence & image processing ,Computer vision ,Artificial intelligence ,Face detection ,business ,Decision model ,Coding (social sciences) - Abstract
To meet the requirement of HEVC conversational video transmission over packet-loss network, a novel error resilient scheme based on region-of-interest (ROI) is proposed. First, the weight value for each coding block tree(CTU) in the face detection region is generated. Then, by using the ROI weight value, motion vector and mode decision method based on weighted distortion methods are developed. At last, the rate control method is introduced at region level and CTU level to avoid an increase in bitrate due to error resilience of ROI. The simulations show that, compared with traditional methods, our scheme improves coding performance of ROI by up to 2.32dB in a packet-loss network.
- Published
- 2017