Particle reinforced polymer matrix composites (PMC) are widely used in the military aviation industry as a stealth coating for incident radar attenuation. These coatings are made up of many layers of different materials with varying thicknesses. In this study, we used terahertz (THz) spectral range by means of time-domain spectroscopy (TDS) to examine the thickness of iron particle reinforced polyurethane Radar Absorbing Paint (FE-PU-RAP) composite coatings non-destructively. For accurate estimation of the individual layer thickness in multi-layer coatings, precise knowledge of the material properties of the individual layers is a prerequisite. Hence, standalone samples of the individual layers have been initially employed to extract the material properties, such as the refractive index, absorption coefficient, etc. Then, the accurate thickness estimation by the time-of-flight measurement principle has been carried out using the reflected THz pulses. Further, since the individual reflected pulses from the multi-layered sample are overlapping in nature due to the optically thin coating thickness, a sparse deconvolution technique has been utilized for extracting each individual reflected time instant. The scattering from the iron particles also provides a challenge to the identification of the interface signals. This was addressed by optimizing the regularization parameter in the sparse deconvolution algorithm. With this technique, accurate estimation of each of the individual layers in the multi-layer sample has been accomplished, otherwise difficult non-destructively.