Research On Image Segmentation Method Based On Asymmetric Geodesic Distance

With the continuous advancement of technology,multimedia data has become a ubiquitous part of people’s lives,especially the amount of image and video data keeps surging.Therefore,image processing has become an extremely critical field.It is not only a content that must be explored in the field of scientific research,but also a basic problem that must be faced and solved in the information age.As an effective image processing method,image segmentation is not only an important part of image understanding and analysis,but also has a wide range of applications in industrial automation,product inspection,character recognition,face recognition,intelligent transportation,and remote sensing satellite image processing and other fields.However,there is no perfect image segmentation algorithm at present.Many problems can affect the segmentation quality,such as long processing time,noise pollution,target occlusion and image shadow.Therefore,it is very important to improve the speed and quality of image segmentation in image processing.Aiming at the problems existing in image segmentation technology,an image segmentation method based on asymmetric geodesic distance is proposed in this thesis.It mainly studies the construction of asymmetric quadratic metric functions,related fastmoving algorithms and key point detection and other problems,and applies them for image segmentation.The main work and contributions are as follows:(1)An image segmentation method based on an asymmetric geodesic growth model is proposed,which can automatically detect a series of new key points,and calculate the optimal connection curve between two adjacent key points until a closed contour is detected.The anisotropic and asymmetric Randers metric function is also used in the key point detection process,and a new key point detection method and stopping criterion based on the asymmetric metric function are proposed to produce more accurate and robust segmentation in different scenarios split results.Compared with the classical key point detection method of anisotropic asymmetric metric function introduced in this study can overcome the problem that the curve has nothing to do with the direction of motion,and obtain more accurate segmentation results.(2)An interactive image segmentation method based on asymmetrical geodesic distance is proposed.The asymmetric quadratic metric function is adopted in the curve evolution model based on Voronoi diagram,and the seed point set located in the foreground and background is introduced into the model.Therefore,the proposed Voronoi graph curve evolution model can integrate the expected evolution direction of curves,artificial interaction information and effective image features,so as to obtain more accurate image segmentation results.Compared with the classical Voronoi graph model based on the isotropic Riemannian metric function,the model can overcome the local extreme value problem that the evolution curve is prone to fall into errors,so as to produce more accurate segmentation results in different segmentation scenarios.In this scheme,the seed point set is combined with the active contour for interactive segmentation,so as to achieve better contour evolution and get better segmentation results.