We report on the method developed by Zibetti, Charlot & Rix (2009) to construct resolved stellar mass maps of galaxies from optical and NIR imaging. Accurate pixel-by-pixel colour information (specifically g – i and i – H) is converted into stellar mass-to-light ratios with typical accuracy of 30%, based on median likelihoods derived from a Monte Carlo library of 50,000 stellar population synthesis models that include dust and updated TP-AGB phase prescriptions. Hence, surface mass densities are computed. In a pilot study, we analyze 9 galaxies spanning a broad range of morphologies. Among the main results, we find that: i) galaxies appear much smoother in stellar mass maps than at any optical or NIR wavelength; ii) total stellar mass estimates based on unresolved photometry are biased low with respect to the integral of resolved stellar mass maps, by up to 40%, due to dust obscured regions being under-represented in global colours; iii) within a galaxy, on local scales colours correlate with surface stellar mass density; iv) the slope and tightness of this correlation reflect/depend on the morphology of the galaxy.