Existing compressed sensing (CS) MRI reconstruction techniques can handle a wide variety of undersampled k-space acquisitions corrupted with noise. On top of that, these methods can also be employed for resolution enhancement applications. Very often, clinicians will ‘zoom in’ on an MR image, where the zooming is actually implicitly performed by bilinear, bicubic or sinc interpolation. These algorithms are computationally simple, but give rise to artifacts. As such, in this abstract, we propose a CS MRI reconstruction algorithm, which uses a sparsifying transformation, tailored to the problem of reconstructing an enlarged image from undersampled k-space data. We will also discuss how we accelerate this algorithm by optimally choosing the free parameters in the augmented Lagrangian reconstruction formulation.