語系:
繁體中文
English
說明(常見問題)
回圖書館首頁
手機版館藏查詢
登入
回首頁
切換:
標籤
|
MARC模式
|
ISBD
Spatiotemporal motion analysis using...
~
University of Southern California.
FindBook
Google Book
Amazon
博客來
Spatiotemporal motion analysis using five-dimensional tensor voting.
紀錄類型:
書目-電子資源 : Monograph/item
正題名/作者:
Spatiotemporal motion analysis using five-dimensional tensor voting./
作者:
Min, Changki.
面頁冊數:
128 p.
附註:
Adviser: Gerard Medioni.
Contained By:
Dissertation Abstracts International67-10B.
標題:
Computer Science. -
電子資源:
http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3238348
ISBN:
9780542928222
Spatiotemporal motion analysis using five-dimensional tensor voting.
Min, Changki.
Spatiotemporal motion analysis using five-dimensional tensor voting.
- 128 p.
Adviser: Gerard Medioni.
Thesis (Ph.D.)--University of Southern California, 2006.
This work presents a novel spatiotemporal approach to produce motion segmentation and dense temporal trajectories from an image sequence. A common representation of image sequences is a 3D spatiotemporal volume, ( x,y,t), and its corresponding mathematical formalism is the fiber bundle. However, enforcing the spatiotemporal smooth motion constraint which is our only motion model is difficult in the fiber bundle representation. Thus, we convert the representation into a new 5D space (x,y,t,v x,vy) which consists of image domain (x,y ), time domain (t), and velocity domain ( vx,vy). In the space, each moving object produces a separate 3D layer, and the smoothness constraint is now enforced on the layers by the tensor voting framework. This single spatiotemporal smoothing step solves both correspondence and segmentation simultaneously. The motion segmentation is achieved by identifying those layers, and the dense temporal trajectories are obtained by converting the layers back into the fiber bundle representation. Unlike most other motion analysis approaches, our proposed approach does not make restrictive assumptions about the observed scene or camera motion and is therefore generally applicable.
ISBN: 9780542928222Subjects--Topical Terms:
626642
Computer Science.
Spatiotemporal motion analysis using five-dimensional tensor voting.
LDR
:03171nmm 2200301 a 45
001
865793
005
20100729
008
100729s2006 ||||||||||||||||| ||eng d
020
$a
9780542928222
035
$a
(UMI)AAI3238348
035
$a
AAI3238348
040
$a
UMI
$c
UMI
100
1
$a
Min, Changki.
$3
1034244
245
1 0
$a
Spatiotemporal motion analysis using five-dimensional tensor voting.
300
$a
128 p.
500
$a
Adviser: Gerard Medioni.
500
$a
Source: Dissertation Abstracts International, Volume: 67-10, Section: B, page: 5955.
502
$a
Thesis (Ph.D.)--University of Southern California, 2006.
520
$a
This work presents a novel spatiotemporal approach to produce motion segmentation and dense temporal trajectories from an image sequence. A common representation of image sequences is a 3D spatiotemporal volume, ( x,y,t), and its corresponding mathematical formalism is the fiber bundle. However, enforcing the spatiotemporal smooth motion constraint which is our only motion model is difficult in the fiber bundle representation. Thus, we convert the representation into a new 5D space (x,y,t,v x,vy) which consists of image domain (x,y ), time domain (t), and velocity domain ( vx,vy). In the space, each moving object produces a separate 3D layer, and the smoothness constraint is now enforced on the layers by the tensor voting framework. This single spatiotemporal smoothing step solves both correspondence and segmentation simultaneously. The motion segmentation is achieved by identifying those layers, and the dense temporal trajectories are obtained by converting the layers back into the fiber bundle representation. Unlike most other motion analysis approaches, our proposed approach does not make restrictive assumptions about the observed scene or camera motion and is therefore generally applicable.
520
$a
We proceed to solve several problems that are hard to solve from the video stream directly because of the dense matching and segmentation steps, but are straightforward with our framework. For instance, image mosaics and 3D scene reconstruction problems require some parameter estimation steps that relate two or more images, and they are theoretically well-defined. However, if the scene contains some large scale moving objects or strong parallax, then the parameter estimation steps become difficult due to the mixed motion information. In this case, our two main results, accurate segmentation and correspondence, can greatly simplify the problems and provide stable results.
520
$a
Another contribution of the work is a GPU-based tensor voting implementation. Tensor voting in the 5D space with a large number of tokens is computationally intensive. The new GPU-based voting implementation achieves significant performance improvement over the conventional CPU-based implementation by taking advantage of the parallel architecture of modern GPUs.
590
$a
School code: 0208.
650
4
$a
Computer Science.
$3
626642
650
4
$a
Engineering, Electronics and Electrical.
$3
626636
690
$a
0544
690
$a
0984
710
2
$a
University of Southern California.
$3
700129
773
0
$t
Dissertation Abstracts International
$g
67-10B.
790
$a
0208
790
1 0
$a
Medioni, Gerard,
$e
advisor
791
$a
Ph.D.
792
$a
2006
856
4 0
$u
http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3238348
筆 0 讀者評論
館藏地:
全部
電子資源
出版年:
卷號:
館藏
1 筆 • 頁數 1 •
1
條碼號
典藏地名稱
館藏流通類別
資料類型
索書號
使用類型
借閱狀態
預約狀態
備註欄
附件
W9077912
電子資源
11.線上閱覽_V
電子書
EB W9077912
一般使用(Normal)
在架
0
1 筆 • 頁數 1 •
1
多媒體
評論
新增評論
分享你的心得
Export
取書館
處理中
...
變更密碼
登入