Camera images may be analyzed to identify perceptual control axes for a lighting system. These control axes may correspond to how human users perceive lighting scenes. A camera may capture high-dynamic-range images of a new room under various lighting conditions. Dimensionality-reduction may be applied to the images to create an image-based map. In this map, each datapoint may correspond to one of the images. The map may be transformed to align it with how human users perceive lighting scenes in the new room. The transformed map may be employed as a control space for the lighting system. In some cases, the map is created without gathering user ratings of lighting scenes for the new room. In other cases, creating the map involves gathering user ratings for as few as three lighting scenes for the new room.