Video compressive sensing with over-completed dictionary

Traditional video cameras have problem with preserving patial and temporal resolution synchronously due to hardware factors such as readout and analog-to-digital (A/D) conversion time of sensors. To overcome this problem without more hardware cost, we propose a new video acquisition system which employs compressive sensing theory to improve the imaging efficiency. Compressive Sensing (CS) is an innovative theory which allows us to combine signal acquisition and compression together, thus capturing compressed signal directly. In this paper, we explore the advantage that video volumes could be sparsely represented under over-completed dictionary. With this characteristic, we can reconstruct the original video with far fewer measurements than conventional Nyquist sampling rate. Experimental results validate that we can obtain promising recovery with limited measurement Also, we gain frame rate improvement without spatial resolution reduction.