In order to overcome the shortcomings of state-of the-art visibility detection methods, a novel algorithm was proposed to measure the atmospheric visibility based on intrinsic luminance of road surface using a video camera. First, the luminance and height constant regions of the road were extracted. Then, the apparent luminance of sampled areas was computed, and power functions were built to calculate the intrinsic luminance according to Koschmieder formula and camera calibration technology. Finally, multipoints sampling was conducted to calculate the visibility distance whose the results were clustered. The experimental results show that this algorithm can lower computation costs and achieve a more precise atmospheric visibility.