Informatics Publishing Limited

Abstract

The misty fog and hazy weather condition within the atmosphere can significantly decay the visibility of a scene. This is due to the atmospheric particles that absorb and scatter the light that travels from the scene point to the observer. This  approach propose a simple and fast method for haze removal from a single input image based on image fusion. The method applies both the white balance and global contrast enhancement to the original image respectively; the method employs a bee colony optimization for accurate haze-free results. We acquire the enhanced results by determining the weight sum of the two inputs. To reduce the number of artifacts introduced by the weight maps, our method is designed in a multiscale fashion by using a Laplacian pyramid representation. This method performs in a per-pixel fashion. Experimental results demonstrate that our current method effectively removes haze, even better than the complex state-of-the-art techniques, and is sufficiently for real-time dehazing applications. 

Keywords

Full Text

Abstract

The misty fog and hazy weather condition within the atmosphere can significantly decay the visibility of a scene. This is due to the atmospheric particles that absorb and scatter the light that travels from the scene point to the observer. This approach propose a simple and fast method for haze removal from a single input image based on image fusion. The method applies both the white balance and global contrast enhancement to the original image respectively, then preceding these two images as inputs that can be weighted by some specific maps. We acquire the enhanced results by determining the weight sum of the two inputs. To reduce the number of artifacts introduced by the weight maps, our method is designed in a multiscale fashion by using a Laplacian pyramid representation. This method performs in a per-pixel fashion. Experimental results demonstrate that our current method effectively removes haze, even better than the complex state-of-the-art techniques, and is sufficiently for real-time applications.

Keywords

Full Text