Different kinds of cereal flours submitted to various technological treatments were classified on the basis of their mid-infrared spectra by pattern recognition techniques. Classification in the wavelet domain was achieved by using the wavelet packet transform for efficient pattern recognition (WPTER) algorithm, which allowed singling out the most discriminant spectral regions. Principal component analysis (PCA) on the selected features showed an effective clustering of the analyzed flours. Satisfactory classification models were obtained both on training and test samples. Furthermore, mixtures of varying composition of the studied flours were distributed in the PCA space according to their composition.