Dynamic vision and imaging systems can substantially improve our quality of life. However, key issues that must be addressed in order to deploy these systems are their potential fragility and the need to process vast amounts of information in real time. As we show in this paper, these issues can be addressed by using systems theoretic tools to recast a wide range of problems into a tractable convex optimization form. These ideas are illustrated with several applications including multiframe tracking, motion segmentation, texture analysis/synthesis and video reconstruction and inpainting.