ImageSegmentationReview

Note on the taxonomy

possible criteria:

• number and type of features used (texture, color, optic flow)
• still images vs video sequences
• variational o not
• local or global
• region or boundary based
• shape priors
• bayesian/MAP
• clustering
• active contours, region merging

another criteria

• edge detectors
• 2 regions
• multi region

edge and boundary detectors

• haralick
• canny

region merging and growing

• KMM, multiscale region merging

active contours

boundary based

• GAC

region based

• GAR: deriche
  • cardelino
• active contour without edges: chan vese

level lines

edge based

• cao
• cao muse \& sur
• desolneaux, moisan, morel

region based

• igual, caselles
• cardelino: greedy and multipartition
• burrus
• maximal stable regions: matas

clustering based

• tepper
• tensor voting???

graph cuts

• normalized cuts
• grab cuts

misc

• huttenlocher
• watersheds
• Techreport (im_proc:segmentation:yima:2009:natural_image_segmentation_report) Shankar Rao, Hossein Mobahi, A. Y. S. S. & Ma, Y. Natural Image Segmentation with Adaptive Texture and Boundary Encoding UIUC Technical Report UILU-ENG-09-2211 DC-244, 2009

mean shift

• comaniciu

multiscale segmentation

• arbelaez
• martin

MDL

• leclerc