一种改进的单帧磁共振图像超分辨率算法

An Improved Single-Frame Super-Resolution Algorithm for Magnetic Resonance Image

医学图像处理是图像处理领域的重点和难点,细节丰富的清晰图像有助于协助专家和计算机辅助诊断。针对磁共振医学图像的特点,提出一种结合小波特征和聚类字典的单帧超分辨率重建方法。在训练阶段,首先分别提取低分辨率图像的多尺度小波特征和高分辨率图像的高频特征,将高低分辨率特征图重叠分块,然后利用K均值算法将特征块聚类,使用K奇异值分解分别训练每一类特征块的高低分辨率字典,形成映射关系;在重建阶段,提取低分辨率图像特征块并分类,使用该类字典原子进行重建。最后,引入迭代反投影算法进行后处理,以进一步提高重建质量。实验结果显示,该算法在内部、外部数据集上,视觉和量化指标都有较好表现,并优于同类算法。

Medical image processing is an important and key problem in image processing. High-resolution images with abundant details contribute to assisting physicians and computer aided diagnosis programs. According to the characteristics of magnetic resonance images, we propose a single-frame super-resolution reconstruction method based on wavelet features and clustered dictionaries. In the training phase, the multiscale wavelet features of low-resolution images and all high- frequency components of high-resolution images are extracted, and all of these feature images are overlapping and separated into patches. Then, K-means algorithm is used to cluster feature patches into several classes, for each class a pair of dictionaries is learned using K-singular value decomposition. In the reconstruction phase, each low-resolution patch is classified and sparsely represented with its corresponding dictionary atoms. Iterative back projection is used for post- processing to further improve the reconstruction quality. Experimental results show that the proposed method outperforms other main-stream methods, both visually and quantitatively.