Raman spectroscopy has demonstrated great potential in clinical and biomedical applications. However, slow data acquisition due to weak Raman signals has prevented its wide use, especially in an imaging setup. In this paper, we propose a novel solution in which the full Raman spectrum can be rapidly reconstructed from a few narrow-band measurements based on Wiener estimation. This method, using both traditional Wiener estimation and modified Wiener estimation, has been evaluated on Raman spectra collected from live, apoptotic and necrotic leukemia cells with several different types of band-pass filter choices. Two rules of thumb were identified to guide the choice of filters. Our results indicate that the agreement between reconstructed and measured Raman spectra was excellent according to either the mean root mean square error or the classification accuracy. Therefore, Raman reconstruction from narrow-band measurements represents a new direction in fast Raman imaging for the investigation of fast changing phenomena. Copyright © 2013 John Wiley & Sons, Ltd.