Title   name
2007년 KSIAM 봄 학술대회
 
  Title
  Speaker   Hahn, Jooyoung  
  Date 2007-05-25
  Host
  Place KAIST
  File  의 1 번째 Real Media 동영상입니다. 의 1 번째 강연자료입니다.
 
Abstract : A fine segmentation algorithm is proposed for extracting objects in an image, which have bothweak boundaries and highly non-convex shapes. The image has simple background colors orsimple object colors. Two concepts, geometric attraction-driven flow (GADF) and edge-regionsare combined to detect boundaries of objects in a sub-pixel resolution. The main strategy tosegment the boundaries is to construct initial curves close to objects by using edge-regions andthen to make a curve evolution in GADF. Since the initial curves are close to objects regardlessof shapes, highly non-convex shapes are easily detected and dependence on initial curves inboundary-based segmentation algorithms is naturally removed. Weak boundaries are also detectedbecause the orientation of GADF is obtained regardless of the strength of boundaries.For a fine segmentation, we additionally propose a local region competition algorithm to detectperceptible boundaries which are used for the extraction of objects without visual loss of detailedshapes.We have successfully accomplished the fine segmentation of objects from imagestaken in the studio and aphids from images of soybean leaves.