Research On Image Segmentation Algorithms Based On Improved Snake Model

Image segmentation is the key and basic preprocessing step in the whole image processing.Research on related issues has always been a hot topic at home and abroad.At present,there are quite a few effective methods in this respect.However,due to the diversity and complexity of the image structure,image segmentation is still an challenging missions.The image segmentation algorithm based on parametric active contour model(Snake model)combined with the priori knowledge of image and the upper visual nature.It has many advantages,such as easy to implement and obtaining closed curve,which has been widely concerned since the day it was proposed.In this thesis,the image segmentation algorithm based on Snake model is studied.Firstly,the basic theory of image segmentation and several classical methods are introduced.Then the research status of image segmentation algorithm based on Snake model is described and analyzed.Several important improvements Snake model and existing problems are discussed.The main work and conclusions of this thesis are as follows:(1)In order to overcome the issues that the existing image segmentation algorithms based on parametric active contour models that the curve are difficult to converge to small deep-concave boundaries,have narrow range of applications,high noise sensitivity,poor segmentation performance over weak edge.a novel algorithm for image segmentation based on EPGVF Snake model is proposed in this thesis.Firstly,the Laplacian operator with isotropic smoothness is replaced by the new diffusion term.Secondly,the p-Laplacian functional is introduced into the smooth energy term to strengthen the force in the normal direction,which drives the active contours into small and deep concave regions of objects.Finally,this algorithm keeps the external force field parallel to the edge direction through the edge preserving term,which protects the weak edges.The experimental results show that the proposed model doesn't only overcomes the drawbacks of the existing models,possesses better segmentation effect,but also improves the anti-noise performance obviously and locate to corners accurately.(2)The existing GVF Snake model image segmentation algorithms generally difficult to converge to deep concavities,especially for the boundary where external forces are easy to form critical points,such as small deep and “?” shape concavities.In order to solve the problems,an improved image segmentation algorithm based on GVF Snake model was proposed.Firstly,a new diffusion vector was introduced into the smooth energy term to enhance the diffusion ability of external forces.Secondly,the tangential and normal diffusion is controlled by the weight function which changes with the gradient of the edge.Then,the kernel function and edge gradient convolution are used to adjust the external forces so as to make the contour curve converge to the deep concave boundary.Theoretical analysis and experimental results show that,the proposed model not only segments image with special structures such as deep narrow and inner width concavities correctly,but also has strong noise robustness,which can protect the image edge details effectively and has better segmentation effect.