Multi-target Image Segmentation Based On Parameter Active Contour Models

Image segmentation has been a focus and hotspot in the field of imageprocessing,it’s a branch of computer image processing field.There is a lot of researchin the field of daily life and have application value.Parameter active contour model(also known as the snake) is the main method of image segmentation and targettracking.Snake model combined with prior knowledge of the image characteristicsand pending issues, and have highly topology skills,this make it very suitable forimage segmentation and contour extraction.Snake has been widely used in imagesegmentation.This thesis introduces the research background and significance,compared theadvantages and disadvantages of several kinds of classic method of active contourmodel in detail.The thesis focus on image segmentation based on multipletargets.Analyzed the key steps in the multiple target segmentation and made a fullexplanation of existing multi-object segmentation algorithm.The work of this thesis isapplied the deflating parameter active contour model to multi-objectivesegmentation.Thesis first improved parametric active contour energy function andadded the direction coefficients in function,this restricted the contour moved outwardand contour tangent deformation,ensured that the snake shrink to the outline of theinternal and greatly reducing the chances of self-cross in similar side.Then theexistence of self_cross in the shrink process of parametric active contour model isshowed and the influence of it on the evolution of the model is explained. Thesisanalyzed projection method,it detecting and treating the self-cross are basedoverlapping the projection of two non adjacent edges in X axis and Y axis. Thecomputational geometry method is also introduced.This article take the computationalgeometry method to judge winding and split the contour exactly.Finally the thesisintroduced the problem of the uniform distribution of sample points in contours andkeeping the shape in segmenting objects. Ensures that each target has surrounded bythe snake, evolution and get the correct convergence results.The experimental resultsshow that the method is efficient and accurately.