The advantage of nonuniformily correction algorithm based on recursive least squares(RLS) in IRFPA lies in its low computational complexity and memory, requirements, which makes it easy to meet the need of projects. However, RLS is not suitable for the real algorithmic model completely since the noise parameter fitted in RLS exists severe deviations. Thus, a novel nonuniformity correction algorithm based on reeursive mixed least squares (RMLS) was developed. In this new algorithm, RLS was substituted by RMLS for monuniformity correction. This proposed algorithm not only has all the ad- vantages of RLS, but also gets more excellent results in estimation precision and convergent speed. The experimental results also demonstrate the effectiveness of the algorithm.