基于机器学习的图像分割算法研究

Review on machine learning techniques for image segmentation

图像分割是计算机视觉研究中重要的一部分,其主要目的是在图像中将兴趣域目标与背景分割,关系到后续的目标识别、图像理解等操作的准确性。经过几十年的发展,许多优秀的图像分割的方法被提出。机器学习是当今时代的研究热点,基于深度卷积神经网络等机器学习的图像分割研究进展迅速。总结介绍了应用于图像分割的几种典型机器学习方法,分析比较了相关的分割原理步骤、优缺点和发展现状。最后分析了基于机器学习的图像分割算法的发展方向。

Image segmentation can split images into areas,which have different specific properties.Also,image segmentation is an important step for extracting the interesting portion of the images,the accuracy and effectiveness of the subsequent image analysis are directly related to the processing results of image segmentation.There are many kinds of image segmentation algorithms purposed so far.Machine learning is one of the most popular methods so far.Image segmentation algorithms based on machine learning are concerned in a large degree due to their wide range of applications and high performance in image segmentation.Machine learning algorithms are put forward to cluster or classify the data in order to integrate similar parts and distinguish parts different with each other.The application of machine learning in image segmentation is mainly pixel-wised segmentation based on the classes of pixels to achieve the final purposes of image segmentation.The segmentation methods based on machine learning algorithms can be used in many kinds of images and graphs.Few specialized review of the image segmentation algorithm based on machine learning is put forward currently.In this paper,we will introduce the image segmentation methods based on machine learning algorithms,including Clustering,Support vector machine,Convolutional Neural Network,Adaboost and so on.The principles and steps of each machine learning segmentation methods will be described as follows.Then we will analyze the relationship between the various algorithms and the advantages and disadvantages of the machine learning segmentation methods mentioned in this paper.Last,the development and the trend of image segmentation algorithms based on machine learning will be analyzed.