1 / 19

Multiobjective Approaches in Image Segmentation

Multiobjective Approaches in Image Segmentation. Ruxandra Cohal. Keywords. Image segmentation. Multiobjective Optimization. Image analysis. Thresholding. Pixel classification. Image segmentation techniques results http://www.cs.toronto.edu/~jepson/csc2503/segmentation.pdf.

jorn
Télécharger la présentation

Multiobjective Approaches in Image Segmentation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Multiobjective Approaches in Image Segmentation RuxandraCohal

  2. Keywords Image segmentation Multiobjective Optimization Image analysis Thresholding Pixel classification

  3. Image segmentation techniques results http://www.cs.toronto.edu/~jepson/csc2503/segmentation.pdf

  4. Image segmentation • image partitioning • similarity/dissimilarity criteria • challenging task of image processing • influence over the result of image analysis

  5. Multipleobjective Optimization • single objective optimization problem • goal: minimize, maximize, achieve a certain value • multiobjective optimization problem • goal: optimize in the same time a certain number of criteria

  6. Multipleobjective Optimization • multiobjective optimization criteria examples: • minimize overall deviation • maximize connectivity • minimize the number of features • minimize the error rate of the classifier

  7. Multiobjective Optimization Methods • Reduce the MO function to a single-objective function • assign a numerical weight to each objective • combine the values of the weighted criteria into a single value

  8. Multiobjective Optimization Methods 2. Simultaneous optimization of several objectives • Pareto approach • dominance relation • Pareto-optimal set = a set of solutions that are non-dominated with respect to each other • Pareto optimal solution sets are preferred to single solutions

  9. Image segmentation techniques • Image thresholding • region segmentation method • popular approach due to its straightforward implementation • challenge: automatic fitting of threshold

  10. Image segmentation techniques • Image thresholding • MO approach: combination of segmentation objectives of two thresholding techniques: • Otsu’s method • Gaussian curve fitting method • the objective functions are optimized using: • weighted sum of within-class criterion • overall probability of error criterion

  11. Image segmentation techniques • Pixel classification - supervised classification • training set of data • classifier • number of classes • numerical characteristics (mean, variance etc.)

  12. Image segmentation techniques • Pixel classification - supervised classification • MO optimization methods can be used to design classifiers • the objective functions are usually related to: • rules definition of the classifiers • error rate of the classifier • diversity measurement

  13. Image segmentation techniques • Pixel classification - unsupervised classification • clustering in image segmentation = the process of identifying groups of similar image primitives • image primitives: • pixels • regions • line elements

  14. Image segmentation techniques • Pixel classification - unsupervised classification • the lack of a precise definition for what a cluster is result in a large number of clustering algorithms • clustering algorithms can find structures (partitions) at different levels of refinement

  15. Image segmentation techniques • Pixel classification - unsupervised classification • from MO perspective, the optimization/search techniques used are divided into • deterministic: exhaustive enumeration • stochastic: generate a near-optimal partition reasonably quickly

  16. Image segmentation techniques • Pixel classification - unsupervised classification • stochastic search technique: evolutionary computation approaches • chromosomes(candidate solutions): coordinates of centers of the partitions • evolutionary operators: selection, recombination, mutation • fitness function: objective function • multiobjective clustering algorithms: • NSGA (2 objectives), NSGA-II(3 objectives) • PESA(2 objectives)

  17. Image segmentation techniques • Pixel classification - unsupervised classification • other approaches • PSO (Particle Swarm Optimization) • SA (Simulated Annealing) • Hill Climbing and Genetic Algorithms

  18. Conclusions • Image segmentation • complex process • involves a set of objective to be optimized • the multipleobjective optimization approach facilitate the use of many types of algorithms

  19. Bibliography • Multiobjective Optimization Approaches in Image Segmentation – The Directions and Challenges – Bong Chin-Wei, MandavaRajeswari

More Related