When lidar is used to detect atmospheric visibility, if cloud, fog, smoke, dust or the hard target exist in the detection path, the atmospheric extinction coefficient will change significantly in the local area, then the lidar return signals occur mutate against the original attenuation trend along with distance. Affected by it, using existing algorithm directly will cause the accuracy of visibility inversion low or wrong inversion. A combination visibility inversion algorithm including breakpoint location, extinction coefficient boundary value determination, extinction coefficient iterative inversion was proposed. Firstly, the breakpoint signal was found and located; Secondly, breakpoint was excluded by using slope method to acquire extinction coefficient boundary value; Finally, atmospheric extinction coefficient and visibility based on Fernald algorithm were inverted. According to the simulation experiments of two kinds of typical atmospheric extinction model, the accuracy of visibility inversion is improved, the more accurate global visibility is obtained. The validity of the proposed algorithm is also been verified by using real return signals of the lidar visibility meter developed by ourselves.