Image segmentation - multiscale energy-based level sets 1.0

This toolbox implements an energy-based segmentation algorithm that usesfinite-difference based level set evolution. (See [1], [2]).Multiscale control of the curve evolution is implemented -- this greatly reducesthe sensitivity to initialization as well as improves the execution speedby factors of 10-1000 and even more([3]).Given a segmentation problem, one can use this GUI in order to find theproper parameters that will yield the desired partition for the class of images used.The functions in the toolbox can be then re-used for a scripting algorithm, if desired.NOTE:This is not a full multigrid algorithm, the scale navigation is provided in 'manual'mode, for "image learning". Once you get the hang of it, you can then code thatbit in. On the same line, the error is calculated and shown, but no decision istaken, the iteration control is also left on "manual".For more details see the references at the end.At the beginning of your session run once the ini.m file, for proper paths, then segm_energy_LS.m.From the GUI open your data using the Load/Save menu, then choose segmentation parameters - Mu : the weight of the separation curve length: the higher its value, the shorter (thus smoother) the length. - Lambda+ / Lambda- : the weights of the In/Out square error terms. - dt: the artificial time step; reduce it when you notice instabilities, can be increased for smooth segmentations with often resets. - Nu : area term, set mostly at zero by most people :-)Start iterating at the roughest scale. You can prolongate towards finer scales and restrictback towards rougher scales using the provided controls. At every step current solution isshown, while under the 'View' menu there are more goodies -- error evolution, error functiongradient, residue, etc.The level set function can be saved for further comparisons, etc.For "straightforward" images of say 300 pixels try smoothness parameters of0.01 - 0.1 and go from there. If the contours start oscillating play with thetime parameter dT and/or with the curve redistancing function.You can set various options:INITALIZATION of the levelset, which can be - arbitrary (circles) - manually drawnREDISTANCING of the levelset - bwdist (fastest) - if not available, use the built-in function (slower)PROLONGATION / RESTRICTION can be cell-based or vertex basedREGULARIZATION - 'atan', best :-) - 'sine', narrowband, for educational purposes onlySCALE DEPTH (default 3) - after altering this field the image scales are automatically recalculatedThe read_me.txt file contains all this info plus more. (click on Help in GUI)References:-----------[1] R. J. LeVeque - "Finite Difference Methodsfor Differential Equations" September, 2005[2] T. Chan and L. Vese, "Active contours without edges"IEEE Trans.Img.Proc., vol. 10, pp. 266d-deOCt277, 2001.[3] W. L. Briggs, V. Emden Henson, S. F. McCormick"A Multigrid Tutorial", SIAM, Philadelphia, 2001

Requirements:	No special requirements
Platforms:	Matlab
Keyword:	Andor,  Contours,  Initalization,  Levelset,  Options,  Oscillating,  Parameter,  Redistancing,  Start
Users rating:	0/10