In this thesis, we illustrate a nonlinear regression method “Gauss-Newton” and an adaptive weighted matrix for color reproduction in digital camera. We aim to reduce the chrominance error in nonlinear domain and thus require computing on the tendency to approach minimum. The 24 color patches in GretagMacbeth ColorChecker are the reference for color fitting. Having precise color representation is crucial in many applications such as medical imaging, press, textile industry and so on. With proper calibration, the color can be subjectively or objectively corrected depends on the target application.