Active Contour Method Based On Local Fitting And Edge Entropy

The Active Contour Model is one of the most often used image segmentation algorithms.Image segmentation is an important study subject in the fields of image analysis,pattern recognition,and computer vision.The active contour image segmentation approach was proposed by M.Kass et al.in 1987,and it has since become a current scientific research boom in the field of image segmentation.The most widely used picture segmentation method that employs variational concepts is the active contour model.Its main benefit is that it can provide a continuous and smooth closed segmentation border even when there is a lot of noise.According to the different ways of energy function construction,active contour models can be mainly classified into two categories: edge-based and region-based,while some researchers have also proposed active contour models based on a combination of edge and region.The majority of known region-based active contour approaches are based on feature similarity between pixel points on the active contour and the associated image region,or on the probability construction of pixel points on the active contour in the corresponding region.Individual pixels or global neighborhood information are frequently utilized to analyze the properties of pixels on the curve when calculating similarity or likelihood.In fact,singlepixel data makes the curve extremely sensitive to noise and texture,but using global neighborhood information frequently contains a lot of irrelevant data,resulting in erroneous boundary localization.To address the issue of being sensitive to noise based on single pixel data,as well as the issue of covering a huge amount of irrelevant data based on global data.In this paper,the active sub-neighborhood active contour model as well as the local edge entropy active contour model are constructed,and the detailed work is as follows:First,in order to solve the problem that the classical active contour model uses singlepixel data leading to curves sensitive to features such as noise and texture,an active contour model based on sub-regions is proposed,i.e.,the active contour model is divided into two sub-neighborhoods,the internal features of the two sub-neighborhoods are considered separately,and the features are brought into the active contour model based on the region,and the active contour model framework based on the active sub-neighborhood model is constructed.Through example analysis,it is verified that the method is applicable to almost all region-based activity profile models and is compatible with the level set method.The experimental analysis verifies that the method has robustness to noise and texture,and also improves the accuracy of boundary recovery.Then,to deal with the problem of covering a large amount of irrelevant information based on global information,the article constructs a local edge entropy activity profile model.The model considers the local ring structure instead of global information,which can effectively avoid the complex amount of information generated by redundant or irrelevant information.Meanwhile,as the active contour approaches the target region gradually,an adaptive local neighborhood radius is designed,which can prevent the active contour model from falling into local minima and stopping the evolution.Moreover,the adaptive local neighborhood radius has the effect of accelerating the iteration.An edge entropy is constructed to determine whether the active contour reaches the edge or not,and the image entropy value is large in the uniform region and small in the non-uniform region.By comparing the existing methods,it is verified that the local edge entropy active contour model method has good segmentation effect on images with uneven brightness.In this paper,two novel active contour models are developed to solve the segmentation problem of complex content images.This will help to advance the active contour research and apply it to more practical problems.