A new energy-dependent multi-cell detector, which is a generalization of the conventional front-back detector, was studied using computer simulations. The noise performance of the detector for bone quantitation was examined in comparison to an ideal energy discriminating detector, and front-back detectors with and without inter-detector filters. The front-back detectors were optimized for a reference object composed of water and bone, and then compared to the new detector over a range of object compositions. In this paper, precision in calculated bone thickness is used as the criterion for evaluating detector performance. Simulations show that the segmented detector always performs better than the front-back detector without an inter-detector filter. It outperforms the detector incorporating a filter by an amount that depends on the heterogeneity of the x-ray spectrum. In addition, for single component radiographic images, this multi-cell detector retains information which is lost in the front-back detector with a filter layer.