Abstract This article presents an approach to reduce the number of color patches printed during a calibration for electrophotography. During a calibration, primary color patches at different halftone levels are printed on a belt and are measured using on-board sensors. Regression models are used to predict primary color tone values on output media from these on-board sensor measurements. Then, the necessary adjustments using tone correction or other control variables are applied to compensate for tone deviation. Laying down and measuring color patches consumes time and toner. It is desired to reduce the number of calibration color patches while ensuring regression model accuracy. A procedure for color patch reduction is proposed in this work. In this procedure, a minimal set of color patches and their associated halftone levels are identified using variable selection algorithms. The results demonstrate that the minimal set contains 47.1% fewer calibration color patches than a full set, while the regression model accuracy of the minimal set is effectively the same as that of the full set.