영상내의 잡음특성에 따른 적응적 쿼드트리 영상분할방법

Metadata Downloads
Issued Date
Now many image segmentation methods have been studied for extracting the object with the special means but it is difficult to extract only the accurate parameters which can distinguish between images and noises on the noise images. So these methods can not be effective segmentation against the noise images. Therefore it has a problem about the potential decreasing of the performance according to noises for all applications using the present quadtree segmentation.
In this paper, we propose an adaptive quadtree segmentation by using the variation and the quartile deviation of the pixels, which can recognize the noises properties and segment effectively the images information on noise images.
Therefore our method can be applied for the fields of the various images processing because it has an advantage to distinguish only an image information from the noise images. As the result of our simulation, we confirm that the proposed quadtree segmentation is more efficient than the present quadtree methods when tested on Gaussian noise and impulsive images.
Alternative Title
An adaptive quadtree image segmentation method reflecting the properties of noises in images
조선대학교 대학원
일반대학원 전자공학과
Awarded Date
Table Of Contents
목차 = i
Ⅰ. 서론 = 1
Ⅱ. 잡음특성에 따른 영상처리의 필요성 = 4
Ⅲ. 영상분할의 이론적 배경 = 8
A. 영역 성장법(Region Growing) = 8
B. 영역 분할법(Region Spliting) = 10
C. 영역 분리와 통합법(Split and Merge) = 12
D. 쿼드트리 영상 분할 = 13
E. 범위(Range)기반 쿼드트리 영상분할 = 17
Ⅳ. 잡음특성에 따른 적응적 쿼드트리 영상 분할방법 제안 = 18
A. 잡음의 종류 = 18
1. 가우시안 잡음(Gaussian noise) = 18
2. 임펄스 잡음(Impulsive noise) = 21
B. 잡음특성 인식 = 21
1. 잡음인식 과정 = 21
2. 이산 푸리에 변환 = 23
C. 화소의 분산(Variance)을 이용한 영상분할방법 = 34
D. 화소의 사분위편차(Quartile Deviation)를 이용한 영상분할방법 = 36
Ⅴ. 시뮬레이션 및 결과 분석 = 38
A. 가우시안 잡음의 경우 = 40
B. 임펄스 잡음의 경우 = 66
Ⅵ. 결론 = 91
참고문헌 = 93
김상진. (2005). 영상내의 잡음특성에 따른 적응적 쿼드트리 영상분할방법.
Appears in Collections:
General Graduate School > 4. Theses(Ph.D)
Authorize & License
  • AuthorizeOpen
  • Embargo2009-12-10
Files in This Item:

Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.