Level Set Theory And Its Application In Image Segmentation

Image segmentation is a pre-processing step for many tasks such as image recognition and target tracking.It plays an important role in image processing and it can be used for various aspects related to computer vision such as medical detection and traffic control.However,there has been no general image segmentation method due to the particularity and diversity of the image data to be segmented.At the same time,maintaining the balance between the accuracy of image segmentation and the implementation of the algorithm is also a problem to be considered in the research process.Due to its flexible topological structure,the level set method has obtained excellent segmentation results in the field of image segmentation.In this paper,level set method and the application of level set theory in image segmentation are deeply studied.The main research contents are as follows: Firstly,the principle method of level set theory,the knowledge of curve evolution theory and the numerical calculation process when level set method is implemented are introduced.The objective evaluation criteria of image segmentation is also introduced to provide the objective basis for the experimental analysis in this paper.Then,the three classical level set models of CV model,DRLSE model and SBGFRLS model are introduced.The theoretical basis of each model is expounded respectively.The experiments of these three level set models are carried out,and the segmentation effects,advantages and disadvantages are summarized and analyzed.Secondly,aiming at the problem that the intensity inhomogeneity in the gray image affects the image segmentation result,a level set gray image segmentation model combined with probability knowledge is proposed.For gray images with intensity inhomogeneity,the bias field function is used to describe the uneven components of the image.And the fitting function is used to represent the intensity of the pixel points inside and outside the image contour,which is more suitable for the original image;the data energy function is designed based on the single point distribution probability of the image to improve the efficiency of the algorithm and make the level set function reach the target position faster.By comparing with the experimental results of LBF model,LS-ACM model,RD model and GDRLSE1 model,we can see that the proposed algorithm overcomes the problem of intensity inhomogeneity effectively and achieves high segmentation accuracy.And through the objective comparison of the two evaluation indicators and running time,the proposed algorithm has good implementation.Finally,the segmentation method of complex color images with color information is studied.In order to obtain a good segmentation effect,a level set color image segmentation model combining saliency features is proposed.The level set function is evolved in multidimensional space to avoid the step of converting color image into gray image,so that retaining more image information and it is beneficial to obtain more accurate segmentation results.At the same time,the energy functional is designed in combination with the saliency features of the images,so that the images are segmented using differences of saliency features in different regions of the image.By comparing the experimental segmentation results and running time of CV model and SBCRDLS model,we can see that the proposed algorithm has good segmentation performance for color images and does not require much computation time.So it is an effective color image segmentation model.