Melt ponds are observed regularly on the surface of the Arctic sea ice in spring and summer. The appearance of these melting features reduces strongly the surface albedo and accelerates the decay of the Arctic sea ice. Until now, the spatial extent of melt ponds on the Arctic sea ice has been explored only sporadically by field campaigns. Although the knowledge of melt-pond distribution on the entire Arctic sea ice would be a solid basis for the parameterization of melt ponds in existing sea-ice models. Owing to the different spectral properties of snow, ice and water a multispectral sensor like the Landsat-7 ETM+ is applicable for the analysis of the distribution of melt ponds. The very high spatial resolution of 30 m is an additional advantage. In the present study we developed an algorithm, based on a principal components analysis (PCA) of two spectral channels, for the determination of the melt-pond fraction. The melt-pond fraction is the ratio of ponded area to sea-ice coverage. The PCA allows, considering two tiepoints, a proper differentiation of melt ponds and other surface types like snow, ice or water. For this analysis, the spectral band 1 with a central wavelength at 480 nm and band 4 with a central wavelength at 770 nm are used as they represent the extreme values of the spectral albedo of melt ponds. The comparison of the melt-pond fraction determined with the PCA method with a method developed by Markus and others in 2003 yields a slightly different spatial distribution of the ponded areas: the fraction calculated with the PCA in areas with an apparently high melt-pond distribution is more compact and areas dominated by open water show a lower fraction than the results of the method from Markus and others, 2003. The overall melt-pond fraction of the two methods results in a correlation coefficient of r = 0.95 for the selected Landsat-7 scene of Baffin Bay from 27 June 2000. The analysis of several scenes through the melting period indicates that the tiepoint values are varying according to the solar incidence and elevation angle. Comparing scenes from different melting stages helps us to develop a procedure finding the optimal value of the tiepoint. This allows an automatic processing routine to evaluate all suitable Landsat scenes.