The authors present an improved iterative method for deriving temperature corrections in stellar atmospheres. At each step of iteration, the procedure computes, from the current values of the radiation field, a set of depth-dependent factors, averaged over angles and frequencies. Even a crude initial estimate of the radiation field leads to a fairly accurate determination of these factors on the first iteration: they are quasi-invariant in subsequent iterations. For a range of effective temperatures the procedure quickly converges to the correct solution of the radiative transfer equation. At each iteration, the temperature distribution is directly derived via the energy conservation constraint.