This paper addresses the problem of generating a super-resolution (SR) image from a single low-resolution input image using compressed sensing. The authors view the low-resolution image as a downsampled version of a high-resolution image, where the patches of the high-resolution image are assumed to have a sparse representation in an over-complete dictionary of prototype signal-atoms. The principle of compressed sensing ensures that the sparse representation can be correctly recovered from the downsampled signal under mild conditions. The paper demonstrates the effectiveness of sparsity as a prior for regularizing the ill-posed SR problem. It also shows that a small set of randomly chosen raw patches from training images can serve as a good dictionary, leading to competitive or superior results compared to other SR methods. The method involves two main steps: first, finding the sparse representation for each local patch using a dictionary of low-resolution patches, and second, refining the entire image using a reconstruction constraint. The local model uses a one-pass algorithm to enforce compatibility between adjacent patches, and the global model ensures consistency with the low-resolution input. The entire process is summarized in Algorithm 1. Experimental results on generic and animal images show that the proposed method outperforms existing methods in terms of edge sharpness and texture clarity. The paper also discusses the preparation of dictionaries from random raw patches and the use of derivative features for low-resolution patch representation. Future work includes determining the optimal number of raw sample patches and integrating dictionaries for images with multiple textures or object categories.This paper addresses the problem of generating a super-resolution (SR) image from a single low-resolution input image using compressed sensing. The authors view the low-resolution image as a downsampled version of a high-resolution image, where the patches of the high-resolution image are assumed to have a sparse representation in an over-complete dictionary of prototype signal-atoms. The principle of compressed sensing ensures that the sparse representation can be correctly recovered from the downsampled signal under mild conditions. The paper demonstrates the effectiveness of sparsity as a prior for regularizing the ill-posed SR problem. It also shows that a small set of randomly chosen raw patches from training images can serve as a good dictionary, leading to competitive or superior results compared to other SR methods. The method involves two main steps: first, finding the sparse representation for each local patch using a dictionary of low-resolution patches, and second, refining the entire image using a reconstruction constraint. The local model uses a one-pass algorithm to enforce compatibility between adjacent patches, and the global model ensures consistency with the low-resolution input. The entire process is summarized in Algorithm 1. Experimental results on generic and animal images show that the proposed method outperforms existing methods in terms of edge sharpness and texture clarity. The paper also discusses the preparation of dictionaries from random raw patches and the use of derivative features for low-resolution patch representation. Future work includes determining the optimal number of raw sample patches and integrating dictionaries for images with multiple textures or object categories.