NIGHT IMAGE DEHAZING ALGORITHM BASED ON DUAL-DOMAIN TRANSMITTANCE ESTIMATION

To solve color distortion and halo issues in the dehazing light source region of nighttime images, a nighttime image dehazing algorithm based on dual-domain transmittance estimation is proposed. Based on channel difference and brightness information, the image was divided into light source region and non-light source region. We obtained the initial transmittance of light source regions and non-light source regions respectively through designed adaptive boundary constraints and dark channel prior. A multi-scale fusion strategy was adopted to reconstruct the dual-domain transmittance, which was refined through guided filter to obtain final transmittance, and the haze-free image was restored. This algorithm can effectively maintain the original color and structural characteristics of light source region. Compared with other methods, the BRISQUE index, FADE index, and NIQE index have all achieved significant improvements.