Image And Text Segmentation For One-dimensional Medical Images And Its Application In Paper Electrocardiogram

How to realize the image-text segmentation for one-dimensional medical imaging is the main focus of this article.There are relatively few related researches on two-dimensional medical image segmentation and one-dimensional content algorithms.Therefore,the study of image and text segmentation algorithms for one-dimensional medical images is essential to promote the automatic extraction of data sets and subsequent related research.Two algorithms are proposed around the unsupervised learning field and the supervised learning field to realize the segmentation of images and texts for one-dimensional medical images.The specific work and innovations are as follows:First of all,in the field of traditional unsupervised learning,this article proposes a curve seam-carving algorithm for one-dimensional medical image segmentation based on the traditional image scaling algorithm seam-carving algorithm.It defines the seam map and the energy map to segment the ECG curve.In addition,in order to ensure the accuracy of the pixel points selected by the algorithm when the step size is large.Secondly,in the field of supervised learning,in view of the shortcomings of the original UNET neural network and the inadaptability of the data set,this article designs two attention modules: new S&C module and skip-attention module.The new S&C module composition can be understood as a parallel connection of spatial attention and channel attention.The purpose is to use the attention mechanism to increase the weight of the target information and suppress the influence of irrelevant information.In addition,the skip-attention module increases the weight of the target information in the content of the downsampling layer by participating in the global information of the feature map of the upsampling layer.Experiments show that curve seam-carving has good image segmentation ability on lowprecision images,but it has the disadvantage of insufficient generalization ability in fractal graphics.The improved UNET model solves two problems: First,it solves the deficiencies of the curve seam-carving algorithm in fractal graph segmentation.Secondly,it solves the problem that the original UNET model does not adapt to the data set.