Optimization of the point spread function by means of sensor-less adaptive optics, based on direct imaging of the focal spot, suffers from errors due to enormous dynamic range of the focal intensity. Also, optimization algorithms based on the focal spot metrics only, are insensitive to other system parameters and can converge to” wrong” solutions. To improve the beam quality and the robustness of the global extremum, we have introduced dynamic feedback control of the camera sensitivity. To further increase the robustness of optimization, we introduced a regularization parameter in the form of some function of the system state, achieving its minimum together with the desired solution. Significant gain in achievable beam quality is shown in comparison with the implementation lacking those improvements. Proposed techniques are implemented in BeamTuner software for fine-tuning of laser and imaging systems with adaptive optics. © 2014 SPIE.