Superresolution Reconstruction of Multiframe Images Using Regularization with a Quadratic Form Observation Model

Two aspects of the superresolution reconstruction algorithm proposed by Fryer and McIntosh (FM algorithm) are considered in this paper. One is the sensitivity to the noise existed in low resolution images, and the other is the high computational complexity in the superresolution reconstruction process. The QM-SRR algorithm, presented in this paper, reconstructs the observation model using matrix form instead of vector form to represent images. Furthermore, total variation is added as a regularization term to improve the noise sensitivity of FM algorithm. Performance analysis and simulation results show that QM-SRR algorithm has very low complexity, can remove the noise efficiently and quickly reconstruct a high resolution image from multiframe low resolution images quickly, is well suited for real-time applications.