The FOCal Underdetermined System Solver (FOCUSS) was originally designed to obtain sparse solutions from limited data. This paper adapts this algorithm for dynamic MR imaging reconstruction from angularly undersampled k-t data. FOCUSS is very effective for dynamic MR imaging such as cardiac imaging because the periodic motion is sparse on x-f space, and the initial low resolution image can be obtained from over-sampled center region in k-space. Experimental results for dynamic MRI from radial k-space data prove that FOCUSS is a very powerful algorithm that reduces angular aliasing artifacts without sacrificing the resolution.