An Improved Dark Channel Prior Method for Dusty Image Restoration

Abstract

For the images which were obtained in a dusty environment, because of the tiny particles suspended in the atmosphere absorb and scatter the light, there is serious haze in the images. The haze can directly cause colour changes and the loss of image details. As a result, a lot of efficient features in the dusty image were concealed or weakened by the haze, which created difficulties to obtain a natural and clear restored image, and has brought serious impact for aerospace, automatic driving and traffic monitoring in the dusty environment. An improved restoration method based on dark channel prior theory and gradient domain guided filtering was proposed to remove the haze and restore the dusty image. According to the characteristics of the dusty images, a specific guided image was adopted to make gradient domain guided filtering for the dark channel image firstly, so that the transmittance image can be better refined than that in the dark channel prior method, and then the image can be restored according to the image degradation model. Experiments show that the visual effect of the restored image is natural and clear, and can retain most effective features, which lays a foundation for computer vision-based target monitoring and recognition in the dusty environment.