北京生物医学工程

图像域自适应线性自回归MRI图像客观质量提升算法研究

An in-planeobjective quality enhancement algorithm of 3D MRI via linear AR model

作者：武博李文志张旭段娟孙智梁志远张楠

单位：首都医科大学生物医学工程学院(北京100069)

分类号：R318.04

出版年·卷·期（页码）：0·0·0（0-0）

摘要：

目的磁共振成像(magnetic resonance imaging, MRI)是临床医学影像检查的重要手段，然而受限于检查时间，高客观质量的MRI图像难以获得，有效提高MRI图像分辨率成为研究热点。本文为高效地提高MRI图像分辨率，减少患者检查时间，提出图像域的基于线性自回归模型自适应客观质量提升算法。方法首先，创建基于距离权重的自回归模型获得插值系数；其次，利用自适应算法解决自回归算法中插值矩阵病态的问题；最后，通过对标准数据库中具有不同噪声水平的图像进行测试。结果本算法重建图像的峰值信噪比（peak signal to noise ratio, PSNR）平均值比经典算法提高1.8~4dB，结构相似度（structural similarity index measurement, SSIM）指标平均提高0.02~0.06。结论图像域自适应线性自回归MRI图像客观质量提升算法在有效提高MRI图像分辨率的同时，保持了较高的客观质量。

Objective Magnetic resonance imaging (MRI) is an important examination in clinical medical imaging. Objective quality enhancement (OBE) is a hot research topic and the aim is to effectively improve the resolution of the MRI. Methods First, an autoregressive interpolation algorithm based on the distance weighting factor is proposed. Second, considering the case of the ill-conditioned matrix, an adaptive interpolation method is adopted. Finally, the simulation is carried out on the brain MRI images with different noise levels from a standard matrix. Results Compared with other classic interpolation methods, the average peak signal to noise ratio (PSNR) of super resolution MRI obtained by this method improves 1.8dB to 4dB and the average structural similarity index measurement (SSIM) improves 0.02 to 0.06. Conclusions The proposed algorithm can effectively improve resolution of MRI and achieve a promising objective quality.

参考文献：

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录