Root morphology is a key trait for plants’ response, i.e. growth and development under various environmental stresses. Traditionally, the plant biologists relied on manual or semi-automatic methods to accurately compute these root morphological traits. Recently, the high-throughput acquisition of root image data permits automatic extraction of root traits. In this context, root segmentation is a key computer-vision task. Current deep learning-based methods proposed for this task are limited due to the diverse root characteristics such as orientation, size, shape and varying degrees of self-occlusion. To overcome these limitations, we propose a novel framework that utilises high-resolution root features for root segmentation. Specifically, the super-resolution network is employed as a pre-processing module since roots in the investigated images are thin and blurry. These enhanced root images are then subsequently utilised in a root segmentation network. In this paper, we also highlight the effect of different resolution levels on root segmentation task with an ablation study.