Tomographic images of porous media are complex distributions of linear attenuation coefficients that reflect the combined effects of scanning spatial resolution, photon statistical measurement errors, and true material densities. We address how the true voxel-scale attenuation distribution and measurement errors are convoluted to yield measured density frequency distributions. A deconvolution algorithm is demonstrated that uses the measured density frequency distributions and known photon statistical errors to quantify average cross-section volume contents of pure components and a mixed-component phase. The mixed-component phase represents regions where components are intertwined or varied in spaces smaller than the scanning resolution. This approach is applied to a complex core of the Culebra Dolomite Member of the Rustler Formation collected at the Waste Isolation Pilot Plant, near Carlsbad, New Mexico. The methodology provides a quantitative measure of the volume content of gypsum, dolomite, and mixed-components, and heterogeneity in the sample. ¿ 1998 American Geophysical Union