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
A stochastic block matching algorith...
~
Kim, Sungook.
FindBook
Google Book
Amazon
博客來
A stochastic block matching algorithm for motion estimation in video coding.
紀錄類型:
書目-電子資源 : Monograph/item
正題名/作者:
A stochastic block matching algorithm for motion estimation in video coding./
作者:
Kim, Sungook.
面頁冊數:
84 p.
附註:
Source: Dissertation Abstracts International, Volume: 57-05, Section: B, page: 3338.
Contained By:
Dissertation Abstracts International57-05B.
標題:
Engineering, Electronics and Electrical. -
電子資源:
http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=9630762
A stochastic block matching algorithm for motion estimation in video coding.
Kim, Sungook.
A stochastic block matching algorithm for motion estimation in video coding.
- 84 p.
Source: Dissertation Abstracts International, Volume: 57-05, Section: B, page: 3338.
Thesis (Ph.D.)--University of Southern California, 1995.
Research on video compression has been very active in recent years. The success of a compression method relies on the effective removal of spatial and temporal redundancies existing in image sequences. One effective method in removing the temporal redundancy is the block-based motion compensated coding, which is adopted in the MPEG and H.261 video compression standards. This research focuses on two important issues associated with motion estimation, i.e. the speed in motion vector search in encoding and the rate-distortion performance of the decoded video. The basic idea is to exploit the spatial and temporal correlations present in the motion field, and two novel algorithms are proposed. They are called SBMA(T) and SBMA(ST), which stand for the stochastic block matching algorithm using the temporal correlation and the spatial and temporal correlations, respectively. The main feature of the proposed algorithms is that the search window for motion vectors has a shiftable window center and a variable window size. Furthermore, a random search or a new modified 3-step search can be applied to avoid exhaustive search at all positions in a given search window to result in a higher speed-up factor.Subjects--Topical Terms:
626636
Engineering, Electronics and Electrical.
A stochastic block matching algorithm for motion estimation in video coding.
LDR
:04038nmm 2200277 4500
001
1840156
005
20050714101548.5
008
130614s1995 eng d
035
$a
(UnM)AAI9630762
035
$a
AAI9630762
040
$a
UnM
$c
UnM
100
1
$a
Kim, Sungook.
$3
1928504
245
1 2
$a
A stochastic block matching algorithm for motion estimation in video coding.
300
$a
84 p.
500
$a
Source: Dissertation Abstracts International, Volume: 57-05, Section: B, page: 3338.
500
$a
Adviser: C.-C. Jay Kuo.
502
$a
Thesis (Ph.D.)--University of Southern California, 1995.
520
$a
Research on video compression has been very active in recent years. The success of a compression method relies on the effective removal of spatial and temporal redundancies existing in image sequences. One effective method in removing the temporal redundancy is the block-based motion compensated coding, which is adopted in the MPEG and H.261 video compression standards. This research focuses on two important issues associated with motion estimation, i.e. the speed in motion vector search in encoding and the rate-distortion performance of the decoded video. The basic idea is to exploit the spatial and temporal correlations present in the motion field, and two novel algorithms are proposed. They are called SBMA(T) and SBMA(ST), which stand for the stochastic block matching algorithm using the temporal correlation and the spatial and temporal correlations, respectively. The main feature of the proposed algorithms is that the search window for motion vectors has a shiftable window center and a variable window size. Furthermore, a random search or a new modified 3-step search can be applied to avoid exhaustive search at all positions in a given search window to result in a higher speed-up factor.
520
$a
In this work, we first point out that the location which gives the minimum sum of absolute difference (SAD) in block matching does not always give the true motion vectors, since they are often corrupted by noise. We propose a first-order Markovian random process to model the temporal correlation existing in motion vectors. Furthermore, by considering the maximum a posteriori probability (MAP) estimation for a given sequence of images, we define the optimal motion vector sequence to be the one which minimizes the average error probability along the image sequence, and formulate its solution as the shortest path problem. In theory, the shortest path problem can be solved via dynamic programming. However, this method cannot be practically applied since too many image frames have to be stored for processing.
520
$a
To overcome this difficulty, a new fast motion vector estimation algorithm SBMA(T) is proposed, which does not require the storage of many frames. It is shown that SBMA(T) has a speed up factor around 200:1 with respect to the traditional full-search block matching algorithm (FBMA). However, the overall coding gain of SBMA(T) in terms of the rate-distortion tradeoff deteriorates. The observation that motion vectors are also highly correlated in the spatial domain helps to improve SBMA(T). By exploiting both spatial and temporal correlations, another new algorithm called SBMA(ST) is proposed, where a spatio-temporal correlation pattern is used to improve the overall accuracy of estimated motion vectors. Compared to full search, we can achieve the speed up factor ranging from 70 to 150, while keeping the coding gain (i.e. rate-distortion tradeoff) about the same. We develop appropriate interface subroutines and incorporate the new algorithms in existing MPEG1 and H.261 simulation packages, and then perform extensive experiments on real test image sequences to evaluate the performance of the proposed SBMA(T) and SBMA(ST).
590
$a
School code: 0208.
650
4
$a
Engineering, Electronics and Electrical.
$3
626636
690
$a
0544
710
2 0
$a
University of Southern California.
$3
700129
773
0
$t
Dissertation Abstracts International
$g
57-05B.
790
1 0
$a
Kuo, C.-C. Jay,
$e
advisor
790
$a
0208
791
$a
Ph.D.
792
$a
1995
856
4 0
$u
http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=9630762
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
W9189670
電子資源
11.線上閱覽_V
電子書
EB
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