Dual-energy imaging shows increased conspicuity and specificity of lung nodule detection through the removal of undesired contrast resulting from overlying bone structures. We have developed an algorithm that automatically determines the optimal cancellation parameters for a log-subtraction technique for a pair of high- and low-energy images. The core algorithm involves shrinking the data, extracting bone features, extracting salient edge from these bone features, calculating a tissue-cancellation map, computing the maximum-likelihood bone contrast cancellation parameter, and finally, calculating the soft-tissue cancellation parameter using an empirical relationship. We verified the performance of the algorithm using observer studies, in which the value of the tissue-cancellation parameter calculated by the algorithm was compared to the value manually selected by nineteen trained observers. A number of dual-energy images were acquired with a modified GE Revolution XQ/i, flat-panel-detector chest imaging system, using an anthropomorphic phantom. The effects of variables such as patient size, kVp, mAs, lung texture, patient motion, and the presence of foreign objects in field-of-view on algorithmic performance were evaluated. We found that the algorithm-selected parameter values had less variability than those selected by the observers. Furthermore, the algorithm-selected parameter was within the limits of the variability of the observers for all cases.