The performance of dot printers has recently been improved. Images output by dot printers can provide simple, economical medical reference images if important diagnostic information is not lost. We developed an image processing scheme for chest radiographs that employed a dot printer. We used two types of pixel value conversions, a nonlinear pixel value correction using a lookup table and a linear pixel value conversion using histogram analysis. The density distribution of chest radiographs was analyzed and classified into high-density and low-density images. The two types of pixel value conversions were used depending on the density of chest radiographs. Converted pixel value had density characteristics that were adapted to the output image of the dot printer, and thus the density distributions of the lungs of radiographs became comparable. In addition, an adaptive unsharp masking technique with processing parameters optimized for each of the lungs and mediastinum was applied. An ROC study for the detection of lung nodules was carried out to evaluate the performance of dot printer images. The area under the ROC curve (A(z)) for dot printer images was 0.816, while sensitivity and specificity were 77.6% and 75.2%, respectively. The performance indicated the usefulness of our image processing scheme.