Magnetic Resonance Imaging (MRI), is a non-invasive imaging modality that provides high soft-tissue contrast making MRI a valuable clinical tool. MRI offers different contrast images depending on the pulse sequence used, and each contrast image holds different diagnostic value. In most cases, multi-contrast images are acquired to provide complementary diagnostic information to radiologists. However, scan time increases with each contrast resulting in lower throughput. Also, patient discomfort increases with time, making patients more susceptible to motion, resulting in images with motion artefacts. Compressive sensing (CS) is a signal processing technique that allows exact reconstruction of a sparse signal, using fewer measurements. CS exploits compressibility or sparseness of a signal in some transform domain that is orthogonal to data domain. MRI scans data in Fourier domain, orthogonal to image, making MRI a natural match to CS. Although CS has been previously applied to MRI, better sparsifying transformations allow higher quality image reconstructions. This study investigates application of CS to MRI. First, single contrast reconstruction of MRI will be explored, then similarity between different contrast images will be exploited for better sparsifying transformations to reconstruct better images in the presence of low data. Finally, a method for finding the sparsifying transformation along with the sparse signal will be explored. The methods will be compared to the methods exist in the literature.