This paper addresses the problem of automatic detection and classification of road pavement surface cracks by the analysis of images acquired during road pavement surface surveys. In particular, pre-processing techniques and their combination are discussed and evaluated. Local statistics, as the mean and standard deviation of pixel intensities, computed for non-overlapping image regions, are considered. The various pre-processing techniques are combined in different orders, to evaluate their impact in adequately modelling the feature space and thus leading to the best distress detection results. A set of well-know metrics is used for the evaluation, to conclude about the best pre-processing order