東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁 到查詢結果 [ null ]

切換: 標籤 | MARC模式 | ISBD

The extraction, restoration and trac...

Jiang, Nan.

FindBook

Google Book

Amazon

博客來

The extraction, restoration and tracking of image features.

紀錄類型:	書目-語言資料,印刷品 : Monograph/item
正題名/作者:	The extraction, restoration and tracking of image features./
作者:	Jiang, Nan.
面頁冊數:	87 p.
附註:	Source: Dissertation Abstracts International, Volume: 70-05, Section: B, page: 3083.
Contained By:	Dissertation Abstracts International70-05B.
標題:	Engineering, Electronics and Electrical. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3360599
ISBN:	9781109181883

The extraction, restoration and tracking of image features.
Jiang, Nan.

The extraction, restoration and tracking of image features. - 87 p.

Source: Dissertation Abstracts International, Volume: 70-05, Section: B, page: 3083.

Thesis (Ph.D.)--Arizona State University, 2009.

Image features are very important and useful in computer vision applications such as matching, superresolution, retrieval, tracking, and object recognition. The main scheme is to compute the descriptor of a salient area, sometimes around an interest point. This dissertation presents proposed algorithms on extraction, restoration and tracking of image features.

ISBN: 9781109181883Subjects--Topical Terms:

626636
Engineering, Electronics and Electrical.

The extraction, restoration and tracking of image features.
LDR:03588nam 2200277 4500 001 1402107
005 20111028105722.5
008 130515s2009 ||||||||||||||||| ||eng d
020 $a 9781109181883
035 $a (UMI)AAI3360599
035 $a AAI3360599
040 $a UMI $c UMI
100 1 $a Jiang, Nan. $3 1020778
245 1 4 $a The extraction, restoration and tracking of image features.
300 $a 87 p.
500 $a Source: Dissertation Abstracts International, Volume: 70-05, Section: B, page: 3083.
502 $a Thesis (Ph.D.)--Arizona State University, 2009.
520 $a Image features are very important and useful in computer vision applications such as matching, superresolution, retrieval, tracking, and object recognition. The main scheme is to compute the descriptor of a salient area, sometimes around an interest point. This dissertation presents proposed algorithms on extraction, restoration and tracking of image features.
520 $a The first part of this dissertation discusses a nonlinear regression process for preserving image features. Image denoise is an important and challenging task for preserving features. This dissertation proposes a novel denoise method that preserves the image feature against severe noise disturbance. The proposed method uses a combination of a range filtering and an anisotropic edge distance filtering. The edge map after each filtering is used for subsequent steps in which pixel intensity is weighted according to the distance to the closest edge point. The process is carried out in an iterative scheme until the edge map stabilizes. Furthermore, we compare existing denoise algorithms with proposed method based on traditional criterion as well as new measurements metrics, including MSE, UIQI, Edge Change Distance and Local Structure Change measurement.
520 $a The second contribution of this dissertation is a fast feature tracking system. The advantage of this system is that it estimates the homography between consecutive frames on long airborne image sequences. The affine model homography is first approximated and then decomposed into independent camera movement including translation, rotation, and scale, and then is modeled as a discrete-time linear system. The states of the camera movement are estimated using Kalman filter. When the assumption of temporal independence between the streams of state and observation noise are not satisfied, or when the estimated noise covariances deviate significantly from the true values, a recursive least-square filter with a forgetting factor is used in place of a Kalman filter to track the camera motion. The estimated homography is used in combination with the dynamic feature points scheme, in order to avoid the false registrations due to noise, self-resemblance of image structure, or too few feature points. In addition, the estimated homography approach can reduce the computational overhead by down sizing the searching space through finding the matching feature descriptors.
520 $a The appendix of this dissertation presents two specific systems for creating mosaics from a sequence of images. Image registration is the fundamental basis for many problems in computer vision. Both systems use feature-based image registration to build the global motion vectors, and use block matching to reveal the local motion vectors induced by foreground moving objects.
590 $a School code: 0010.
650 4 $a Engineering, Electronics and Electrical. $3 626636
690 $a 0544
710 2 $a Arizona State University. $3 1017445
773 0 $t Dissertation Abstracts International $g 70-05B.
790 $a 0010
791 $a Ph.D.
792 $a 2009
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3360599