Interactive Image Segmentation Method Based On Asymmetric Geodesic

The application of digital image technology has continuously improved people ’s lives and improved production efficiency.As a key technology in digital image processing,image segmentation can separate the target area from the background pixels by identifying and locating the image area.This technology can better filter and extract the required information from the rich image information,and provide an important basis for the subsequent image processing process.The segmentation result directly determines the final image processing effect.Therefore,image segmentation technology has always been a hot spot in the field of digital image processing research.At present,with the increasing amount of digital image data,people ’s demand for efficient and convenient image processing technology is also increasing.However,due to the difference of segmentation effect requirements in different scenes,as well as the influence of noise and complex background,effective and accurate image segmentation methods become the difficulty of research.Therefore,it is of great significance and application value to study stable and accurate segmentation methods in the field of image segmentation.According to the overall research ideas of theoretical research,method design,model test,algorithm application and expansion,this thesis proposes an image segmentation method based on asymmetric quadratic metric function and geodesic voting model.It mainly studies the construction of asymmetric quadratic metric function,geodesic voting theory combined with obstacle space and the setting of characteristic function based on geodesic voting value,and applies it to image segmentation.The main research contents of this thesis are as follows:(1)Research on geodesic voting model based on asymmetric quadratic metric function.The problem of image segmentation based on isotropic metric function without considering image asymmetry and anisotropy is studied.First,for the input image,we allow to add an endpoint set that is adaptively distributed around the closed contour.Secondly,by comparing the known isotropic metric functions,we construct an asymmetric quadratic metric function in the metric space,which combines the minimum path with the asymmetric quadratic metric,allowing the asymmetry and anisotropy of the image edge to be considered in the geodesic path generated from the endpoint to the interaction point.Then,the physical space with obstacles is used to ensure the extraction of the final closed contour.Experiments with isotropic metric function show that the interactive image segmentation model based on asymmetric quadratic metric has excellent characteristics such as accuracy,efficiency and stability.(2)Research on extended application based on geodesic voting value.We incorporate the saliency features of the image into the construction of the asymmetric quadratic metric function.In order to further improve the accuracy of segmentation,we use the voting value as a new image saliency feature to construct the potential function for the connection of geodesic paths.Firstly,the high voting value contrast at the edge of the image can make the geodesic path more suitable for the image contour,reduce the occurrence of shortcut problems from the model function construction,and increase the robustness of the model.Finally,we compare the voting value feature function with the experimental results based on the image gradient potential function to verify the ability of the model for image segmentation.A theoretical extension of image segmentation combined with allowable path is proposed.