To investigate the possibility to characterize radioactive wastes using available information and measurements in a coherent frame, a Bayesian formalism is built to couple gamma-ray spectrometry and tomographic scans. The gamma ray spectrometry is performed with four High Purity Germanium (HPGe) detectors placed around the waste and scanning 2 cm thick slices of a radioactive waste drum in a Gamma Scanning mode. The tomography provides the density of the drum matrix and identifies heterogeneities. The unknown chemical composition of the matrix and heterogeneities is handled with two parameters that represent the mass fractions of carbon and hydrogen of a fictive {C; Sn; H} material that would show a gamma ray attenuation behavior similar to that of the true material. The approach is tested by simulating the measurement of 239Pu gamma rays produced by a PuO2 sphere placed in a drum slice in which two distinctive zones have been identified by tomography, in presence of 137Cs background. With the studied simulated examples, depending on the sphere position, the use of the density prior information allows decreasing the plutonium mass uncertainty by up to a factor 3 compared to the case when the density prior is not used. In addition, the use of the density prior information provides a posterior plutonium mass distribution having a significant number of event related to the true plutonium mass, which is not the case when the density prior information is not used.