An improved visibility restoration of single haze images for security surveillance systems

With the wide range of features and charms, security surveillance systems are nowadays collective in most industries around the globe. These applications can range from mugging and destruction deterrence to traffic and weather monitoring and more. The surveillance systems are major part in investigations related to crimes and all. But the core problem with the pictures taken with the surveillance system is, it holds many atmospheric particles such as mist, fog, haze, dust etc. By the presence of these particles, the visibility of the images are unfair which poses difficulty in analyzing the image. Image degradation can cause problems for many systems that must operate under a wide range of weather conditions. As such, researchers have been employing various visibility restoration methods to reduce/remove the degradation that has happened while capturing the image. Haze removal is one such visibility restoration practice, which is considered as one of the main reason that is cause degradation in images. Hence, this research emphases on the image degradation that may occur due to haze. Haze is conventionally an atmospheric phenomenon where smoke, dust and other dry particles obscure the transparency of the sky. The dehazing of image is considered quite challenging as the concentration of haze is different for different places, and hence detecting and measuring them is purportedly the greatest challenge. The phenomenon of Color attenuation rollout in images allows us to set the color and intensity of light. Hence, this phenomenon can be used to define the attenuation and hence, can be well used to find the depth of the haze in the images. This research proposes to use the color attenuation prior for single image dehazing. This simple and powerful prior can help to generate a linear model for the scene depth of the hazy image. By learning the parameters of the linear model with a supervised learning method, the link between the hazy image and its corresponding depth map is built effectively. It is thus proposed, to use depth information recovered from the images, to remove the haze from the images.