An adaptive anchored neighborhood regression method for medical image enhancement

Abstract

Chinese Government has launched ambitious healthcare reform aiming to provide better healthcare services for both urban and rural residents via remote diagnosis. Recently, the requirement of high-resolution (HR) images becomes more urgent in the medical field, especially for remote diagnosis. Remote diagnosis is an important means of the Internet of Medical Things (IoMT), which senses patients’ health status according to the medical images and transfers clinical data. Owing to the superiority of reconstruction speed and quality, adjusted anchored neighborhood regression has attracted much attention. However, the hypercells formed by neighborhoods may not center on atoms, and hence it is not accurate to group the neighborhood centered on atoms. In this paper, we propose an adaptive medical image enhancement method. Specifically, we cluster training samples into neighborhoods centered on patches. The LR space is re-divided by replacing dictionary atoms with cluster patch centers as the center of hypercells. By this means, the neighborhood anchored in the patch is defined by computing its K nearest patches, and then applied to pre-compute the projection to map low-resolution patch onto the HR domain. Average quantitative results show that our method is superior to the compared methods in two common medical datasets by 1.19% and 2.32%, respectively, while maintaining the similar running time.