A polarization-based smoke removal method for surgical images

An improved smoke removal model for surgery is introduced with transmission parameters related to a medium’s optical depth rather than scene distance. Theoretical analysis and observation of experimental data shows that cross-polarized signals generated by multiple scattering are less affected by smoke compared to co-polarized signals. We analyze the transmission process of linearly polarized light interacting with different media, and then use polarization difference imaging and color channel information to detect smoke and estimate the transmission parameters. Several further processing procedures including parameter compensation and image smoothing are implemented to recover tissue visibility from surgical images.