In this work, a new method of analyzing noninvasive reflection spectra is presented. The approach explicitly models the inhomogeneity of chromophore distributions in living tissues and thus extracts not only apparent chromophore concentrations but also relative chromophore distributions in tissues. Furthermore, it works with spectra obtained with short source-detector separations where the diffusion theory of light transport through turbid media is not valid, and formerly presented methods thus fail. The effect of inhomogeneously distributed chromophores in a multicompartment model of tissues on measured reflection spectra is explained and an algorithm to deconvolute tissue spectra based on this model is presented. It is evaluated using simulated spectra and measurements on phantoms, which are made up of partially printed pieces of paper to simulate inhomogeneous dye distributions. Its applicability to real tissue is proven using reflection spectra obtained with 130 microm source-detector separation from a hemoperfusion stop experiment. The proposed model accurately determines apparent chromophore concentrations and corresponding distributions in simulated spectra and phantoms. Regarding real tissue spectra, the results correspond to former publications and the spectral reconstruction yields only minimal residuals, indicating a complete and accurate spectral deconvolution. In conclusion, the presented approach is a suitable extension and amendment to existing models of light transport through inhomogeneous samples.