Image Processing - Segmentation Variational Models Osher-Sethian Level-Set Framework + topologically...

15
Regional image segmentation using B-Spline level-set function Author: Gábor Bernát Scientific advisor : d r. ing. Szilágyi László

Transcript of Image Processing - Segmentation Variational Models Osher-Sethian Level-Set Framework + topologically...

Regional image segmentation using B-Spline level-set functionAuthor: Gbor BerntScientific advisor: dr. ing. Szilgyi Lszl

1The Problem

What do you think when you see the following images:Image Segmentation with problems like:

The segmentation itselfMultiple independent segmentsHigh noise levelsLow contrastScalability (rigidity of the contour)2The SolutionImage Processing - SegmentationVariational ModelsOsher-SethianLevel-Set Framework+ topologically flexibleChan-VeseRegional Segmentation+ can handle noise+ low contrastActive ContoursMichael UnserB-Splines as Filters+ scalability+ easy implementationSignal ProcessingOlivier Bernard: Variational B-Spline Level SetIEEE Transactions On Image Processing, June 20093How it works?

The Level-Set Function:

The evolution of the level-set:4Brief Algorithm Break Down

5ImplementationObject-oriented programming C++Enclose functions and variables into a class: BSplineLevelSetSimplify the usage for the user

6UsageFrom an application:

7Graphical User Interface

8PerformanceImage (size)The scale (step size) of the discrete B-Spline function1248Time(s)Number of iterationsTime(s)Number of iterations.Time(s)Number of iterationsTime(s)Number of iterationsLeaf (128128)0.24000070.22500370.24084780.2105877Gradient (128128)0.360040120.265840100.271019100.2508747Spiral (128128)0.647694240.520128160.304954110.2018917Stems (128128)0.436916150.273539100.25297790.29739511Brain MRI (128128)0.24731480.20042370.16722760.1800826Echo Image (387387)4.001256153.128002112.20984781.9428367Leaf Oliver Bernard (2009) Intel 1.4 GHz CPU + 1 GB RAMScale 4 2.03 second Intel 2.53 GHz CPU + 4GB RAM9Key milestone check #1

Results:Run-time sneak peek:10Key milestone check #2

11Other areas where we can use it

12To where from here?Implementation on GPUImprove level-set minimization timeAdapt it to various medical image problemsTracking the left ventricle in an echocardiographic videosSegmentation of teeth in computer tomography images13Questions??????????????????

14Thank you for your attention!

15