KSC 2019
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
On study of Superpixel based approach for Dynamic Mode Decomposition in video processing |
¿µ¹®Á¦¸ñ(English Title) |
On study of Superpixel based approach for Dynamic Mode Decomposition in video processing |
ÀúÀÚ(Author) |
Ngo Thien Thu
Md Alamgir Hossain
Md Imtiaz Hossa
Sung Yun Woo
Eui-Nam Huh
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 46 NO. 02 PP. 1165 ~ 1167 (2019. 12) |
Çѱ۳»¿ë (Korean Abstract) |
|
¿µ¹®³»¿ë (English Abstract) |
Dynamic Mode Decomposition (DMD) is an emerging method that is built based on the power of Singular Value Decomposition (SVD). DMD outperforms Robust Principle Components Analysis (RPCA) in terms of computational performance. By decomposing a complex system into low ranks and sparse, DMD shows that it is a very powerful tool to analyze the underlying spatial-temporal coherence of non-linear systems. When applying to video processing, the DMD makes a new way of video representation by taking each frame as a snapshot of data in time. In this paper, we study a new approach to boost the computational performance of DMD for video processing using superpixels. Unlike the original DMD based on the whole frame, we take into account the diversity of the homogeneous region to improve the computation time. The experiments on different categories of video show that the superpixels approach can reduce the computation time of the original algorithm by more than 50 percent.
|
Å°¿öµå(Keyword) |
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|