DEEP MR IMAGE SUPER-RESOLUTION USING STRUCTURAL PRIORS.
DEEP MR IMAGE SUPER-RESOLUTION USING STRUCTURAL PRIORS.
复制标题
使用结构先验的深度 MR 图像超分辨率。
DOI：
10.1109/icip.2018.8451496
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
Monga,Vishal
中科院分区：
文献类型：
--
作者：
Cherukuri,Venkateswararao;Guo,Tiantong;Schiff,StevenJ;Monga,Vishal
高分辨率磁共振（MR）图像对于准确诊断是期望的。实际上，图像分辨率受到硬件、成本和处理限制等因素的限制。最近，深度学习方法已被证明可以为图像超分辨率产生令人信服的最先进的结果。特别注意所需的高分辨率MR图像结构，我们提出了一种新的正则化网络，利用图像先验，即低秩结构和锐度之前，以增强深度MR图像超分辨率。然后，我们的贡献是将这些先验知识以一种易于分析的方式纳入卷积神经网络（CNN）的学习中，以完成超分辨率任务。这对于低秩先验来说特别具有挑战性，因为秩不是图像矩阵（以及网络参数）的可微函数，我们通过追求秩的可微近似来解决这个问题。Sharpness通过Laplacian的方差来强调，我们表明可以通过网络输出端的固定反馈层来实现。在两个公开的MR脑图像数据库上进行的实验显示出有希望的结果，特别是当训练图像是有限的。
High resolution magnetic resonance (MR) images are desired for accurate diagnostics. In practice, image resolution is restricted by factors like hardware, cost and processing constraints. Recently, deep learning methods have been shown to produce compelling state of the art results for image super-resolution. Paying particular attention to desired hi-resolution MR image structure, we propose a new regularized network that exploits image priors, namely a low-rank structure and a sharpness prior to enhance deep MR image superresolution. Our contributions are then incorporating these priors in an analytically tractable fashion in the learning of a convolutional neural network (CNN) that accomplishes the super-resolution task. This is particularly challenging for the low rank prior, since the rank is not a differentiable function of the image matrix (and hence the network parameters), an issue we address by pursuing differentiable approximations of the rank. Sharpness is emphasized by the variance of the Laplacian which we show can be implemented by a fixed feedback layer at the output of the network. Experiments performed on two publicly available MR brain image databases exhibit promising results particularly when training imagery is limited.