This study investigates the correlations between the topography of different damaged rough surfaces and process conditions. Several surfaces are measured and compared to determine if they can be discriminated. The analysis is performed by using Gaussian Filtering, Wavelet Transform and a more recent approach named Discrete Modal Decomposition. Standardized 3D roughness parameters are computed for each multiscale method, filter (e.g., high-pass, low-pass and band-pass) and available scale. The relevance (i.e., the ability to discriminate surface topographies corresponding to different process conditions) is then investigated using a statistical analysis based on the MesRug TM expert system. The results indicate clear differences between the multiscale methods and show that the Wavelet approach is useful when characterizing localized surface defects while Gaussian Filtering is more appropriate for highly periodic morphological structures. For more complex topographies, this study also clearly shows that the Discrete Modal Decomposition exhibits compelling abilities that fall between those of the Gaussian and Wavelet approaches; this method is clearly more relevant than the Gaussian method in the case of localized defects and less relevant in the case of highly periodical structures and fractal surfaces (1=f α spectrum). This can be explained by the modulated frequency/ amplitude descriptors generated via the modal basis.