Breast cancer may manifest as microcalcifications (μCs) in x-ray mammography. However, the detection and visualization of μCs are often obscured by the overlapping tissue structures. The dual-energy subtraction imaging technique offers an alternative approach for imaging and visualizing μCs. With this technique, separate high- and low-energy images are acquired and their differences are used to “cancel” out the background tissue structures. However, the subtraction process could increase the statistical noise level relative to the calcification contrast. Therefore, a key issue with the dual-energy subtraction imaging technique is to weigh the benefit of removing the cluttered background tissue structure over the drawback of reduced signal-to-noise ratio in the subtracted μC images. In this report, a theoretical framework for calculating the (quantum) noise in the subtraction images is developed and the numerical computations are described. We estimate the noise levels in the dual-energy subtraction signals under various imaging conditions, including the x-ray spectra, μC size, tissue composition, and breast thickness. The selection of imaging parameters is optimized to evaluate the feasibility of using a dual-energy subtraction technique for the improved detection and visualization of μCs. We present the results and discuss its dependence on imaging parameters.