Integrating Haze Density Features for Fast Nighttime Image Dehazing

Abstract

To date, much progress has been achieved on daytime image dehazing, yet the nighttime dehazing problem is still not well addressed. Different from the imaging conditions in the daytime, the ambient illumination in the nighttime hazy scene is usually not globally isotropic due to the non-uniform incident lights from multiple artificial light sources. Currently, almost all the existing nighttime dehazing methods use a certain kind of image priors, whereby these spatial filtering based priors are not widely applicable in nighttime hazy scenes. For example, the maximum reflectance prior (MRP) cannot handle the dark regions well and the dark channel prior (DCP) is not valid in the light source regions. In this paper, we propose an efficient and fast method for nighttime image dehazing. By exploring the visual properties of hazy images, we construct an effective linear model to build the connection between the transmission and multiple haze-relevant features. Towards solving this model, a data-driven approach is adopted to learn the unknown coefficients. Operating on the pixel level, this novel approach requires no further refinement of transmission map as used in those prior-based methods. In addition, aiming at solving the problem of halo effect around the light sources caused by MRP, we introduce a color-dependant MRP method for color correction. We demonstrate the effectiveness of our method on a number of experiments compared to the state-of-the-art nighttime dehazing methods.

SIDE—A Unified Framework for Simultaneously Dehazing and Enhancement of Nighttime Hazy Images

“…While most conventional daytime dehazing methods can be applied for scene radiance recovery, He’s dark channel prior (DCP) [ 2 ], Meng’s boundary constraint method (BC) [ 4 ], Color Attenuation Prior (CAP) [ 6 ], and Berman’s non-local method (NL) [ 7 ] are selected for demonstrations. The performance of the proposed SIDE is also compared with several state-of-the-art nighttime dehazing approaches, including Zhang’s Maximum Reflectance Prior (MRP) [ 11 , 12 ], Li’s Glow and Multiple Light Colors (GMLC) [ 13 ], Yu’s Pixel-wise Alpha Blending (PAB) [ 9 ] and Lou’s Haze Density Features (HDF) [ 41 ], where the parameters are set as defined in the references.…”

“…Although PAB [ 9 ] is capable of increasing the contrast of the scene to a certain degree, apparent halo can be observed in Figure 8 e. In addition, the brightness of the result remains dim. While MRP [ 11 ], HDF [ 41 ], GMLC [ 13 ] and the proposed SIDE can significantly increase the visibility and suppress halo effect, the proposed SIDE has better contrast improvement in local regions. Color distortion can be seen in grove regions in MRP [ 11 ] and GMLC [ 13 ], while it is more natural in our result.…”

“…Color distortion can be seen in grove regions in MRP [ 11 ] and GMLC [ 13 ], while it is more natural in our result. HDF [ 41 ] also results in residual haze in the grove region. Although no ground-truth reference image is available, the result of SIDE has the best subjective performance on color constancy.…”

“…In Figure 10 , it is noticed in MRP [ 11 ] result that, over-saturation around lamps can be observed and halos are also significant in the dehazed result. Although haze is removed in PAB [ 9 ] and HDF [ 41 ] results, halo artifacts still exist. Moreover, the result suffers from dim and distorted illumination.…”

“…Although the color constancy is well preserved, halo effect is still obvious due to the use of guided filter. Lou et al constructed a linear model to connect transmission and haze-relevant features and employed a learning approach to solve the model [ 41 ]. Kuanar et al introduced a CNN based DeGlow model with a embedded DeHaze module for nighttime dehazing [ 42 ].…”