A feedback method is proposed for improving the color accuracy of printer calibration. Calibration of color printers involves data transformation between the device-dependent color space (e.g., CMYK) and the device-independent color space (e.g., CIE Lab). Although the goal of printer calibration is to get an accurate inverse transform T⁻¹(Lab) = CMYK, the forward transform T(CMYK) = Lab, also known as the printer model, is often derived first. Once both T and T⁻¹ are optimized, a cascade, consisting of the inverse transform T⁻¹ followed by the forward transform T, can be used to simulate the real printer and estimate the color output L₁a₁b₁ for a desired input Lab. The accuracy of calibration can be evaluated by calculating the difference ΔLab = L₁a₁b₁ – Lab. Since it is reasonable to assume that the difference ΔLab is most likely due to the inaccuracy of the inverse transform T⁻¹, a negative feedback of ΔLab to the original input Lab would give us an improved result of color calibration. Examples of applying the feedback idea to printer calibration with a halftone-algorithm independent printer model, the two-by-two printer model, are presented.