Revisit Retinex Theory: Towards a Lightness-Aware Restorer for Underexposed Images

Figure 1

Pipeline of our lightness-aware underexposed image restorer, LiAR. For the input image , the initial estimation of its illumination map is taken as the maximum of {R, G, B} channels’ intensities. Then, the illumination map of the scene is estimated from by internal optimization under color and texture constraints and . After that, lightness-aware illumination adjustment is applied to to get the corrected illumination map . The final restoration result is obtained as the production of the reflectance and the corrected illumination .