Robust L1 Norm Factorization in the Presence of Outliers and Missing Data by Alternative Convex Programming