In order to produce a higher resolution image from a low-resolution o ne,a machine learning super-resolution image reconstruction method via regularized sparse representat ion is presented.A convex variational model is proposed for image super-resolution with sparse representa tion regularization.We further introduce two adaptive regularization terms into the sparse representation frame work to improve the processing effect.Firstly,the Graph Laplacian based image clustering model which takes the local manifold structure into account to a given patch is selected to regularize the image local structures. Secondly,the image nonlocal self-similarity is introduced as another regularization term,which improves th e quality of reconstructed images. Therefore,the sparse representation of a low-resolution image patch can be app lied with the high-resolution image patch dictionary to generate a high-resolution image patch,and then,the high -resolution image is obtained.Experimental results proved that the proposed alg orithm outperforms some state-of-the-art super-resolution methods both in pe ak signal-to-noise ratio (PSNR) measure and visual qua lity.