閾値法，k-mean法 領域拡張法 SNAKE法

1 1 k-mean SNAKE 2 0501001502002503000123456789x 10-3 Probability density function P P ( ) 3 0501001502002503000123456789x 10-3 Probability density function LL+ + =++=22222221211122112)(exp212)(exp21)|() |()( fPfPfpPfpPfp :},..,{},,{:,..,221121 PP 4K-means )0(1m)0(2mf)(fh )1(Tf)(fh )0(1m)0(2mf)(fh)0( T)0(T 5 K-means 6 (region growing )Seed point seed point 7 ( )

2 CT 8 ( ) Yan Kang, Klaus Engelke, and Willi A. Kalender, A New Accurate and Precise 3-D Segmentation Method for Skeletal Structures in Volumetric CT Data, IEEE Trans. Med. Im., , , MAY 20039 Snake Stammberger, Magnetic Resonance Imaging, 17, 7,pp. 1033-1042, 19991988 Kass 10 ()[]))1()0(( 1,0,)(),()(uuu= =ssysxs +=1022int))()(()(dssusuuEsss =10ext))(()(dssuPuE))((IsuIP =)))(())(((GsuIsuGP = 2))((EesudP = )()(extintsnakeuuEEE+=EGI))((PPPsuP++=)( sus)(suss ,,,,))((suG ))((suI))((sud)(su Snake min Membrane energy Thin-plate energy =222)(exp21),( xxGM.

3 Kass et al, Snakes: active contour models, Intern. J. Computer Vision, 1, 4, (1988)11 +=1022int))()(()(dsssEsssuuu Snake =sysxss,)(u()[]))1()0(( 1,0,)(),()(uuu= =ssysxs)(su222)( + =sysxssus s: =2222,)(sysxsssu =2222,)(sysxsssu =sysxss,)(u 102)(dsssu 102)(dssssu 12 =10ext))(()(dssuPuE))((IsuIP =)))(())(((GsuIsuGP = 2))((EesudP = EGI))((PPPsuP++=))((suG ))((suI))((sud =222)(exp21),( xxGDOG (difference of Gaussian)

4 D13 MRI 14 Noise ( ) Pardo*, Carreira, A. Mosquera, D. Cabello Asnake for CT image segmentation integrating region and edge information , Image and Vision Computing 19 (2001) 461-475