Image Dehazing by Image Enhancement and Multi-scale Laplacian Pyramid Fusion

Abstract

When atmospheric particles, such as dust, smoke or water droplets, accumulate in the air, a haze phenomenon occurs. The inclement weather conditions diminish the visibility, reduce the contrast and change the color of an image. The poor visibility may cause a serious problem to various computer vision applications such as object recognition in surveillance systems, transportation systems, etc. Therefore, image dehazing is an essential task, used to remove the influence of weather factors from the image. In this paper, a novel single image dehazing method is introduced which improves the visibility and contrast of the hazy image. To achieve a haze-free image, the original hazy image is processed through basic image enhancement-based operations such as log transformation, inverse log transformation and guided filtering. The resulting set of multi-exposure images is then merged into a single high-quality haze-free image through a multi-scale Laplacian fusion. The experimental evaluation is performed on a large set of challenging hazy images in terms of both qualitative and quantitative parameters. The results obtained through the proposed fusion-based method not only remove the haze effectively but also able to highlight the micro details of the objects in the image.