The digital camera has become an integral part of our everyday life, thus rendering image processing and correction a popular field of research. Corrections of red-eye, colors, blur and focus – are already an integral basic element in any serious camera. Scenes with haze, fog or smoke provide a special challenge due to the inhomogeneousness of the phenomena. Current solutions mostly require the use of at least couple of images, making the solution impractical for daily use.
In this project, we will introduce a statistic prior, the Dark Channel (1. He Kaiming, Sun Jian and Tang Xiaoou, 2008), as a means to perform de-hazing based on a single image and a physical model of the phenomena. The algorithm offers a robust solution, yielding impressive results for a variety of situations: scenes in panoramic, urban and indoor locations, as well as aerial photography.
