Orthogonal Matching Pursuit with Noisy and Missing Data: Low and High Dimensional Results