This study investigates techniques for accounting for the nonlinearity of the input/output response of a camera system. A simple power-law form of the nonlinearity was assumed and estimates of the value for the exponent for each of the color channels were made using three different methods. The responses from an Agfa StudioCam camera were linearized and then device characterization was attempted. Characterization errors were up to 10% better using the spectral-sensitivities-based method for estimating the nature of the nonlinearity than using the other two methods. We therefore suggest that the spectral-sensitivities-based method should be preferred for characterization or any other computational process that requires linearization of the camera responses. We expect greater benefits using this method for "low-end" camera systems and/or for cameras where the spectral sensitivities are known or more precisely estimated We also expect the smoothness of the illumination to influence the error in the estimates of the nonlinearity using the luminance- and mean-reflectance based methods. (C) 2004 Wiley Periodicals, Inc.