An algorithm for segmentation of diesel spray images has been developed. Its most important feature is robustness against experimental setups that fail to guarantee images whose histograms show two distinct peaks. According to the approach presented, only the peak from the background is used, and it is assumed that the background peak is narrow enough not to include too much of the spray. The algorithm has proved successful for evaluation of images from a pressurized vessel as well as from an engine with optical access, with no need for adjusting the tuning parameters. By adjusting them, one may tune the noise sensitivity.