Morphological image segmentation for realistic image representation preserving semantic object shapes의미객체 형태를 보존하는 영상표현을 위한 형태론적 영상분할

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 353
  • Download : 0
As an attempt to achieve realistic image representation by segmentation, a morphological image segmentation scheme is proposed. The proposed scheme mainly aims to represent homogeneous visual objects with few regions while preserving semantic object shapes as faithfully as possible. This strategy is based on the fact that homogeneous visual objects occupy most parts of the entire image domain in a typical ````head and shoulder```` video sequence. For realistic or precise segmentation, a bottom-up approach has been reported to be more appropriate than a top-down method. A bottom-up approach begins with initial image segmentation. Since, all subsequent processes strongly depend on initial segmentation, it should provide sufficient information for an image efficiently. Acquiring such initial segmentation, an efficient marker extraction algorithm is proposed. It first identifies candidate areas where effective markers can exist, and transition areas are excluded from marker extraction process. We regard them as marker clusters and propose a multi-valued marker extraction algorithm from marker clusters by the local histogram analysis in each marker cluster. Then, we perform region classification by thresholding the morphological gradient image and refining the class perimeters by utilizing initial segmentation. An ordered and classified region-merging algorithm is suggested to reduce the number of regions within classes. This algorithm considers only regions having the same class identifier and provides self-extracted threshold values for region merging from the class area itself. At the end, we eliminate redundant small regions deliberately by considering their neighborhood. We take the size of a small region as the macroblock size in video compression standards, since it has been regarded large enough to estimate reliable features for a region. Experimental results show that the resultant segmentation can be a well self-contained representation of an image. Tha...
Advisors
Ra, Jong-Beomresearcher나종범researcher
Description
한국과학기술원 : 전기및전자공학과,
Publisher
한국과학기술원
Issue Date
1999
Identifier
156172/325007 / 000935822
Language
eng
Description

학위논문(박사) - 한국과학기술원 : 전기및전자공학과, 1999.8, [ viii, 111 p. ]

Keywords

Marker; Region merging; Image segmentation; Morphology; Region classification; 영역분류; 마커; 영역병합; 영상분할; 형태론

URI
http://hdl.handle.net/10203/36529
Link
http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=156172&flag=dissertation
Appears in Collection
EE-Theses_Ph.D.(박사논문)
Files in This Item
There are no files associated with this item.

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0