Magnetic resonance scans have recently struggled with inherent limitations such as spatial resolution and long examination times. A novel, rapid compressively-sensed magnetic resonance high resolution image resolution algorithm is presented in this paper. By combining a highly sparse sampling scheme and the super-resolution reconstruction (SRR) method, this technique addresses these two critical issues. Due to the extremely difficult requirements for the accuracy of diagnostic image registration, the presented technique makes use of image priors, deblurring, parallel imaging, and discrete dense displacement sampling for the deformable human body and motion analysis. Clinical trials as well as phantom-based research have been carried out. It has been demonstrated that the proposed algorithm can improve image spatial resolution while reducing motion artifacts and scan times. Convolutional neural networks (CNNs) perform admirably when used to reconstruct images obtained by compressed-sensing magnetic resonance imaging (CS-MRI). The latter goal of the research was to improve subimages’ quality by developing a novel iterative reconstruction method that uses image-based CNNs and k-space correction to preserve original k-space data. CNNs represent a priori information about image spaces in the proposed method. The CNNs are first trained to map zero-filling images onto full-sampled images. The zero-filled part of the k-space data is then recovered. Following that, k-space corrections are used to preserve the original k-space data by replacing unfilled regions with original k-space data. The processes described above are used iteratively.