Data analysis of the Observatoire pour la Min¿ralogie, l'Eau, les Glaces, et l'Activit¿ (OMEGA), the imaging spectrometer aboard Mars Express, necessitates the use of techniques able to extract the relevant information of large data sets. Numerous efficient algorithms have already been developed for similar purposes for terrestrial imaging spectrometers such as the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS). We propose here a complementary method based on the wavelet transform. Our algorithm allows the detection of the absorption bands in each spectrum and returns their position, width, and depth. We use it either to study separately the properties of each absorption band or as a classification method, grouping together the pixels whose spectra contain the same absorption bands. We test our algorithm on an AVIRIS observation of Mauna Kea, Hawaii. The method identifies the dominant mineralogy of this area: variability in the ferric and ferrous mineralogy and several types of phyllosilicates. We also study six Martian observation sessions acquired by the Infrared Spectrometer for Mars onboard Phobos 2. We identify ferric and ferrous absorption bands in the data set and variations in the ferrous mineralogy corresponding to different proportions of high and low calcium pyroxene. Finally, we apply our technique to the first 6 months of Mars surface observation by OMEGA/Mars Express, resampled on a 1¿ ¿ 1¿ grid. We thus retrieve the global large-scale variability of the OMEGA data set and identify a global spectral discrepancy inside Martian dark regions which corresponds to the spectral difference between type I and type II terrains previously identified by the Thermal Emission Spectrometer.