摘要
随着临床对医学图像高分辨率的要求,基于低分辨率医学图像的超分辨率重建算法已成为研究热点,该类方法在不需要改进硬件设备的情况下,可以显著提高图像分辨率,因此对其进行综述具有重要意义。针对医学图像领域中特有的超分辨率重建算法,首先分析了该类算法的研究现状,并将其分为三类:基于插值的超分辨率重建、基于重构的超分辨率重建和基于学习的超分辨率重建。同时,基于MR图像、CT图像、超声图像等细分医学图像领域,深入分析了超分辨率重建算法的研究进展,并对不同类型的算法进行了归纳总结和比对分析。其次,对超分辨率重建算法所对应的评价标准也进行了介绍。最后,展望了超分辨率重建技术在医学图像领域的发展趋势。当前应用于医学图像领域的超分辨重建算法已经发展到一定水平,逐步突破基于单一方法的研究形式,通过与机器学习和稀疏表示等理论的深度融合,形成了更高效的算法。
With the demand of high resolution in clinical medical image,the super-resolution reconstruction algorithm based on low-resolution medical images has become a research hotspot. This method can significantly improve the image resolution without the need of upgrading hardware devices,so it is of great significance to review them. In terms of the special super-resolution reconstruction algorithms in the field of medical images,this paper firstly analyzes the research status of related algorithms,then divides them into three categories: the algorithm based on interpolation,reconfiguration algorithms and learning algorithms. Meanwhile,based on the subdivision of medical image field,including MR image,CT image and ultrasonic image,the research progress of super-resolution reconstruction algorithm is deeply analyzed,and different types of algorithms are summarized and compared. Secondly,the evaluation criteria of the super-resolution reconstruction algorithm are also introduced. Finally,the development trend of super-resolution reconstruction technology in the field of medical imaging is prospected. At present,these kinds of technologies have been developed rapidly,and infused with other methods,such as machine learning and sparse representation theories,which may bring out more efficient algorithms.
作者
成云凤
汪伟
CHENG Yunfeng;WANG Wei(University of Shanghai for Science and Technology,Shanghai 200093)
出处
《北京生物医学工程》
2019年第5期535-543,共9页
Beijing Biomedical Engineering
基金
上海市教委2016年青年教师培养资助计划(ZZsl15012)
上海理工大学光电学院教师创新能力建设项目(1000302006)资助
关键词
超分辨率重建
医学图像处理
字典学习
稀疏表示
卷积神经网络
super-resolution reconstruction
medical image
dictionary learning
sparse representation
convolutional neural network
