We propose a quantitative method to evaluate the ability of adaptive image filters for reducing noise to low-contrast objects in Multi-Detector Row Computed Tomography (MDCT). The digital phantom for evaluating the ability of a nonlinear noise reduction filter to a low-contrast object in a MDCT image was produced. The digital phantom consists of a water phantom image of MDCT and a bar pattern with a known contrast and spatial frequency. The edge-preserving adaptive filter for selectively eliminating noise in low-dose scanning known as the Quantum Denoising System developed by Toshiba Medical Systems was chosen as a nonlinear noise reduction filter for this paper. In this paper, two non-linear noise reduction filters, Q1 and Q2 in QDS were employed. By analyzing the power spectrum (PS) of a digital phantom, the filter performance as regards the contrast and spatial frequency of an object was investigated. The contrast-to-noise ratio from PS could express the feature of the filter briefly. That is, the validity of a filter can be shown based on the combined information of the contrast and spatial frequency of an object. Probably, this simple method will contribute to evaluating various noise reduction filters efficiently.