Research On Variational Level Set Image Segmentation Model

Image segmentation is one of basic and important work of image processing.Many scholars have studied it.Among them,variational level set image segmentation models have attracted the most attention of scholars.The variational level set active contour model implements image segmentation by minimizing the energy functional on the level set function,which mainly includes two items:the data fidelity term and the regularization term.Although many achievements have been made in the research of digital fidelity and regular terms,there are still many shortcomings,For example:Robustness in processing intensity inhomogeneity images and noisy images.In order to better handle the above problems,article has made some improvements to the data fidelity and regular terms and the research content is as follows:1.Propose a variational level set model with kernel metric-based local image fitting(KLIF)energy to segment images in the presence of noises and intensity inhomogeneity.Firstly,a kernel local fitting image(KLFI)is introduced by minimizing a local energy with kernel metric-based data term.And then,we construct a variational level set model with local image fitting energy utilizing the kernel local fitting image obtained in the first step,which can be seen as a measurement of the difference between the fitting image and the original image.Furthermore,two regularization terms are employed in the energy to keep the level set function to be stable during the evolution.At last,the fixed-point iterative algorithm and gradient descent with three-step time-splitting scheme are used to alternately update the local fitting image and the level set function,respectively.The experimental results show that the proposed model is effective to segment image with noises and intensity inhomogeneities.Furthermore,compared with several state-of-the-art variational models,the proposed model shows the best performance2.Propose a new region-based active contour image segmentation model which combining the variable coefficient p-Laplace regular term and cosine fitting energy.Firstly,Introduced the cosine fitting energy which can better eliminate the influence of noise and capture the boundary contour of the object.Secondly,the variable exponent p-laplace energy is used for the regularization of the zero level contours that move to the accurate object boundaries with complex pological changes and deepe Experimental results show that the improved model can segment complex and noisy images,and is not sensitive to the initial contour during the evolution process.