With a view to map the health status of mosaics, non-destructive testing methods ought to be used for data collection. Among these, the infrared thermography is highly recommended since it is non-contact, non-intrusive, non-invasive and able to convert the invisible thermal energy, into a video signal, in which the energy level are usually correlated to a color or a gray scale. The need to provide the position of sub-superficial defects in the clear way is of paramount importance when the diagnostician is not the final client. In the cultural heritage field, raw thermograms, sometimes, do not provide interesting results for the restorer, since they are affected by an undesirable content of noise that limits the detection of what is present beneath the surface. In this work, the multi-dimensional ensemble empirical mode decomposition (MEEMD) technique was used – to the best of our knowledge for the first time – as regards to the thermographic diagnosis of mosaics. It seems to overcoming the thermal barrier of the tessellatum layer, composed by aggregates of different natures, as typical in the Roman era. The results obtained after the inspection via a very long pulse are encouraging, above all whether compared with other algorithms applied in the recent past. The use of intelligent sensors placed inside and outside the mosaic sample, which measured the temperature evolution along the heating up and cooling down phases, helped in the understanding the optimal heat flux to be provided.