Research On Image Segmentation Method Combined MMTD And Bionic Algorithm

Image segmentation is a process of dividing image into sub-areas with different characteristicsand extracting interested object. It is the basis of further image processing and analysis. Many ofexisting segmentation methods can produce good segmentation result but with the drawback ofslow speed. It is an important target of image segmentation as well as this paper’s main researchcontents to improve the segmentation speed while maintaining a satisfied segmentation effect.This thesis introduces bionic algorithm Particle Swarm Optimization into image segmentationand proposes MMTD-improved PSO algorithm to solve the prematurity problem in standard PSO.The new algorithm evaluates particles’ fitness value with MMTD which was proposed based onmedium mathematics system and assigns each particle a different inertia weight according to itsfitness value. In this way, the particle group can search in different scales and thus increase thegroup’s diversity. Experimental results show that the proposed algorithm solved the prematureproblem of PSO while improving its performance.This thesis studies the Otsu algorithm and uses MMTDPSO to improve original Otsu’sefficiency which is limited by its exhaustive calculation by considering Otsu’s threshold calculatingprocess as an optimization problem of objective function. Experimental results show that newalgorithm combined Otsu and MMTDPSO can produce similar segmentation result to original Otsualgorithm while improving the segmentation speed.This thesis also studies the FCM algorithm and uses MMTDPSO to improve original FCM’sefficiency which is limited by its large amount of calculation by considering FCM’s clusteringprocess as a problem of minimization. Experimental results show that new algorithm combinedFCM and MMTDPSO can produce similar result to original FCM algorithm while improving thesegmentation speed.