IPIU 2018 (Á¦30ȸ ¿µ»óó¸® ¹× ÀÌÇØ¿¡ °üÇÑ ¿öÅ©¼¥)
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
Learning based MV-HEVC Disparity Compensation |
¿µ¹®Á¦¸ñ(English Title) |
Learning based MV-HEVC Disparity Compensation |
ÀúÀÚ(Author) |
WEI LIU
WEI LI
SUJING PAN
TaeHoon Yoon
SangUn Park
Yong Beom Cho
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 30 NO. 01 PP. P1 ~ 0073 (2018. 02) |
Çѱ۳»¿ë (Korean Abstract) |
|
¿µ¹®³»¿ë (English Abstract) |
This paper presents the use of support vector machine to replace DC (disparity compensation) in multi-view HEVC decoder. MV-HEVC provides support for coding multiple views with inter-layer prediction. It is being designed as a high-level syntax only extension to allow reuse of existing decoder components. In the profile of MV-HEVC decoder, we can see that DC module takes up a large part, especially when the number of cameras increases, and video pixels are high such as 2048¡¿1556 and 1920 x 1080. Based on the above reasons, we are using HEVC features on CTU depth, bit allocation, and motion vector (MV) to predict P view and B view replace disparity compensation to improve MV-HEVC decoding frame rate. The experiment result show that the proposed method improves 1.3 times frame rate compare with HTM 16.0.
|
Å°¿öµå(Keyword) |
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|