For DCE MRI applications, the images of adjacent time frames are often similar, especially when motion is minimal, in which case temporal TV is a reasonable regularization term. Temporal constraint reconstruction (TCR) has been developed to reconstruct dynamic images from undersampled k-t space data based on such prior information. However, the convergence speed of the algorithm highly depends on the initialization method. In this study, we study initialization using a composite high resolution image based on a jigsaw sampling pattern during pre-contrast frames. The proposed initialization method converges much faster than a conventional initialization method using low resolution images, especially at high reduction factors. In vivo breast imaging experiments were carried out to evaluate the performance of the proposed method. Experiments show the new initialization method allows TCR to achieve a high reduction factor up to 40 without compromising much of the spatial or temporal resolution. The reconstruction errors are much lower than those using the low resolution initialization when the same number of measurements is used.