Based On The Narrow Band Level Set Model Face Contour Extraction Research

Face contour can show the basic characteristics of face very well, so it can be appliedto face recognition research. It is one of the key issues to extract face contour in digitalimage processing for the traditional edge contour extraction methods that have theirlimitations which can’t meet the requirement for the contour extraction of a more complexfacial image. In this case a method based on active contour model arises for target contourextraction at the historic moment. Active contour model is divided into two kindsaccording to their expressions: One is parameter active contour model which is also called“Snake” model that can extract the outline of target quickly, but it can’t process thetopological structure changes when the target has concave and convex area and the outlinecan’t be moved closer to target edge. Another is geometric active contour model which isalso called level set method that is different from “Snake” model because curve is basedon geometric measurement parameters to realize the deformation and the face contourextraction result is better.This paper makes a study of face contour extraction based on level set method with“Snake” model and narrow band algorithm. At first it expounds the current domestic andforeign face recognition and face contour extraction research background, history, presentsituation and the significance in detail. Secondly the following issues are introduced indetail: the traditional active contour model, the basic theory of curve evolution theory,level set method basic principle and the main ideas of narrow-band algorithm; Finally,aimed at the extraction of face contour a method based on narrow band level set model isproposed, This algorithm combined with the existing parameter active contour model andgeometric active contour model advantages, reasonably avoiding the shortcomings,making this kind of algorithm have fast convergence rate curve and very good adaptabilityof topology structure changing, that can contour face contour quickly and effectively. Inaddition we make a simulation experiment aiming at narrow band level set model facefeature contour extraction method. The experimental results show that, a model based onthe narrow band level set has actual maneuverability, which can extract face contourquickly and accurately aiming at the most broad and universality in color JPG format faceimages and get very good effect. The extracted face contour line can fully reflect facefacial features to meet the relevant requirement of face recognition, and can be applied to further research of face recognition.