Hyperspectral imaging (HSI) was used to investigate spectroscopic changes in fresh salmon stored under different atmospheres (air, 60% CO2/40% N-2 and 90% vacuum) and to determine whether HSI can classify fillets by the type of packaging. Hyperspectral images of samples kept at 4 degrees C were acquired and bacterial growth and lipid oxidation were measured. Principal component analysis was applied to study spectral development of samples during storage and K nearest-neighbour classifier was used for the classification of fillets by packaging type. Partial least squares regression was run to reduce the number of wavelengths included in the classification model. The results demonstrated that spectral variations could be observed in the different packaging atmospheres primarily at the wavelengths 606 and 636 nm. Using HSI, successful classification of fillets according to the packaging used (>88%) was achieved and this was largely dependent on spectral characteristics at the wavelengths 606 and 636 nm, possibly due to the different oxidation states of the haem proteins in the muscles. (C) 2011 Elsevier Ltd. All rights reserved.