Dynamic MRI is a technique which obtains time series of images at a high frame rate. In dynamic MRI, it is very important to reduce the data acquisition time because it is often not fast enough to meet the Nyquist sampling rate. Downsampling k-t space measurements accelerates the acquisition speed but may incur aliasing artifacts. To resolve this problem, lattice sampling patterns have been widely used. Lattice sampling pattern in k-t space leads to periodic replications of original image in x-f space. There are some algorithms which use this property and exploit spatio-temporal correlations to solve aliasing problem. Recently, they have drawn considerable attention because they successfully reconstructed high spatio-temporal resolution. Meanwhile, after the advent of the ”compressed sensing”, high fidelity of reconstruction is possible from much less number of measurements than Nyquist sampling requirements. However, lattice sampling pattern is not suitable in compressed sensing perspective. In this paper, new algorithm is proposed which combines compressed sensing, lattice sampling pattern, and parallel imaging. The experimental results show high spatio-temporal resolution without aliasing artifacts in reconstructed cardiac cine image. Furthermore, the proposed algorithm outperforms existing method that uses lattice sampling pattern.