基于数字图像连续表示的图像分割方法  

Image segmentation method based on continuous representation of digital image

作　　者：娄联堂[1] 汪然然 LOU Liantang;WANG Ranran(College of Mathematics and Statistics,South-Central Minzu University,Wuhan 430074,China)

机构地区：[1]中南民族大学数学与统计学学院,武汉430074

出　　处：《中南民族大学学报（自然科学版）》2022年第3期374-378,共5页Journal of South-Central University for Nationalities：Natural Science Edition

基　　金：国家自然科学基金资助项目(60975011)。

摘　　要：研究了一种基于数字图像连续表示的图像分割方法.首先根据机器学习模型的性质,将二维图像的分割问题转换为连续泛函的优化问题;其次利用数字图像的连续表示探讨连续泛函的数学表达式,使其能够表示基于深度学习的图像分割过程;接着通过建立连续泛函的约束条件,将优化问题转化为线性方程组求解的问题;最后利用梯度下降求解方程组,以实现复合绝缘子憎水性图像的分割.An image segmentation method based on continuous representation of digital image is studied.Firstly,according to the nature of the machine learning model,the segmentation problem of two-dimensional images is converted into a continuous functional optimization problem.Secondly,the continuous representation of the digital image is used to explore the mathematical expression of continuous functional,so that it can express the image segmentation process based on deep learning.Then by establishing continuous functional constraints,the optimization problem is transformed into a solving problem of linear equation system.Finally,the gradient descent is used to solve the equation system to realize the segmentation of the hydrophobic image of the composite insulator.

关 键 词：机器学习 图像分割 数字图像连续表示 梯度下降 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]