Underwater image restoration method based on polarization characteristics

Underwater polarization imaging is a highly promising image restoration technique. To effectively eliminate backscattered light from polarization images, we employ the Stokes vector to represent polarization information. Building on the premise that backscattered light is partially polarized, we establish an inversion model for underwater imaging and refine the backscattered light estimation method using the dark channel prior theory. This algorithm effectively suppresses backscattered light interference and enhances the transmission of target information light. Experimental results show that, compared to original images, our algorithm significantly improves image restoration quality metrics across varying turbidity levels, target types, and lighting conditions, demonstrating its exceptional performance in underwater image processing.