Based on the imaging principle of cameras, a transformation equation among response value under different capture settings is established, and a colorimetric characterization model based on the homogeneous polynomial regression is proposed, which avoids errors caused by the nonlinear extrapolation in the traditional polynomial model. A Nikon D3x digital camera is employed to conduct the colorimetric characterization experiment under variable capture settings and the effect of different numbers of terms on characterization accuracy is analyzed. The experimental results show that the homogeneous polynomial model is applicable to describing the colorimetric characterization of cameras with varible capture settings, and it shows the best performance with 6 or 7 terms. In the experiment of 28 kinds of LED light resources with 4 kinds of color temperature and 7 illumination levels, but under fixed color temperature and variable illumination, the predicted color accuracy is superior to 1.5 CIELAB units and it is comparable to the accuracy from the polynomial model with fixed parameters and light sources. In contrast, under the simultaneous change of color temperature and illumination levels, the average chromatic aberration is less than 2.0 CIELAB units, which fulfills the colorimetric accuracy requirement in the practical application of cameras.