Taylor dispersion analysis is an absolute and straightforward characterization method that allows determining the diffusion coefficient, or equivalently the hydrodynamic radius, from angstroms to submicron size range. In this work, we investigated the use of the Constrained Regularized Linear Inversion approach as a new data processing method to extract the probability density functions of the diffusion coefficient (or hydrodynamic radius) from experimental taylorgrams. This new approach can be applied to arbitrary polydisperse samples and gives access to the whole diffusion coefficient distributions, thereby significantly enhancing the potentiality of Taylor dispersion analysis. The method was successfully applied to both simulated and real experimental data for solutions of moderately polydisperse polymers and their binary and ternary mixtures. Distributions of diffusion coefficients obtained by this method were favorably compared with those derived from size exclusion chromatography. The influence of the noise of the simulated taylorgrams on the data processing is discussed. Finally, we discuss the ability of the method to correctly resolve bimodal distributions as a function of the relative separation between the two constituent species