Optimization is the process by which a design is improved by changing the values of a set of parameters (called variables) such that the value of a merit function is reduced, or ideally, driven to zero. The process requires definition of performance criteria via the merit function and effective variables to achieve this goal. This article provides a recommended approach for the optimization specifically of non-sequential optical systems. The recommended approach is:

• Use Pixel Interpolation on all detectors that are used in the merit function, to avoid quantization effects due to pixelated detectors.
• Use aggregate data from these detectors, such as RMS spot size, RMS angular width, angular centroid, centroid location etc rather than the data contained in specific pixels. This 'Moment of Illumination' data is much smoother to optimize than any specific pixel's data.
• Use the Orthogonal Descent optimizer at the start of the optimization process, and then refine the result with the damped least squares and/or Hammer optimizers. Othogonal Descent is usually much faster than the damped least squares optimizer, but may yield slightly less optimal solutions. Use Orthogonal Descent first.

As an example, a free-form mirror is optimized to maximize the brightness of an LED from 23 Cd to >250 Cd in just a few minutes of optimization.