Nighttime visibility meter based on digital photography is an instrument for visibility observation at night with the principle of man’s eyeballing. The visibility is acquired by the luminance information of target light source image collected by CCD, and the key technology is the accuracy of splitting target source and blackbody from the collected image. A light source image processing method based on control tags watershed segmentation approach is introduced. Experimental results show that the watershed method has a good effect on extracting the target light source luminance, and can satisfy the requirements of visibility observation at night. A good agreement has achieved through the correlation analysis to Vaisala FD12 forward scattering instrument.