Abstract The detection of thermal bridges in buildings is one of the key points to be taken into account in energy saving procedures during refurbishment works. Passive infrared thermography (IRT) has been applied for years to detect thermal bridges by referring to the International Organization for Standardization (ISO) 6781:1983. However, the successfulness of this norm is strongly affected by the detection accuracy of the thermal imprint produced on the facade by a conductive material called as “defect” in this work. The drop shadow effect, also produced by the surrounding environment on the facade under inspection, plays indeed an important role during the defect evaluation procedure since it can mask/modify the natural thermal evolution due to diffusion. Many real-life signals acting in the space physics domain exhibit variations across different temporal scales. This work presents an application of a new multiscale data analysis method, the Iterative Filtering (IF), which allows to describe the multiscale nature of an electromagnetic signal working in the long-wave infrared (LWIR) region. IF appears to be a promising method minimizing the influence of the shadows projected on the facade under inspection; subsequently, it allows the optimization of the detection of thermal bridges via sparse principal component thermography (SPCT) technique. The latter inherits the advantages of PCT allowing more flexibility by introducing a penalization term. Here is shown how the accuracy of the defect detection increases after the application of the IF mathematical procedure. Results are discussed on the basis of a couple of case studies referring to dissimilar buildings. Finally, a signal-to-noise-ratio (SNR) comparison with raw data is added to the discussion of the results.