Image quality and contrast become worse in poor visibility condition such as low-light and bad weather.
Since this contrast degradation negatively impacts image understanding,
it is necessary to improve the image quality before the captured images are used in computer vision applications.
Various visibility restoration algorithms have been proposed in order to capture as many features as possible in poor conditions.
One approach is to use an image formation model and to improve image quality by learning parameters of the model,
for example, image degradation factor in low-light condition or light transmission factor in the presence of fog.
Recently, Dong et al. found connectivity of two degradation models in low-light and fog by observing statistics on the negative image of a nighttime image.
Similarily, inspired by the dark channel assumption, we proposed a two-step tone mapping method to enhance contrast of nighttime video.
Images captured at nighttime is prone to noise with lower signal power, and denoising is required to improve quality of nighttime images.
In order to speed up denoising algorihtm for video, we proposed a region-based smoothing filter.
Y. Gong, Y. Lee, and T. Nguyen, "Nighttime image enhancement applying dark channel prior to raw data from camera",
ISOCC, 2016. [link]