Research On Image Segmentation Based On Neural Network And Active Contour

In a rapidly-developing contemporary society,images have become one of the most important ways to quickly obtain information.Due to the continuous advancement of science and technology.Image segmentation,as one of the most intractable aspects of image processing,it's also one of the most important aspects of life,work,and study.The traditional image segmentation is mainly divided into two areas: regional segmentation and boundary segmentation.In the neural networks that have emerged in recent years,image segmentation is divided into semantic segmentation and instance segmentation.In this paper,the deep neural network is used to segment the image and obtain the approximate location of the target.As the initial position of the boundary segmentation,an accurate segmentation result is obtained.The main neural network used in medical image segmentation is called U-net neural network.The shape resembles the English letter U.The left side uses a convolutional layer to learn image features.At the same time,the downsampling layer is used to increase the receptive field,and the right side uses upsampling.At the same time as the layer,copy the image feature on the left and perform new feature learning.Because the downsampling layer will destroy the detailed information of the image,this paper proposes an extended convolutional neural network without loss of image information and which can increase the receptive field.The main idea is to use the extended convolution instead of the downsampling layer to segment the right ventricle of the heart,and to copy the information of the previous layer in the middle layer so that the network can learn more image information.Due to the accuracy of medical images,we use U-net neural network,improved Unet neural network and threshold segmentation method to perform rough region segmentation,giving the approximate position of the right ventricle of the heart,and providing the initial position for the following precise edge segmentation.We propose an N-C-V segmentation model in the boundary distribution of images.The boundary detection function is added to the model to make the model have a better effect in edge segmentation.In order to simplify the calculation and eliminate the reinitialization step,we use the variational level set method when numerically solving the N-C-V model.In the experiment,we compare the model results of the N-C-V model with the traditional Snake model and the Level-Set level set method and DRLSE model.