Image Dehazing Using Superpixel-Guided Algorithm Adaptively for Non-Homogeneous Haze

Abstract

An emerging technique in image processing field is the image dehazing which has a significant influence on various imaging tasks of high level. Although many of the current image dehazing methods are intended for homogeneous haze, but the haze is typically non-homogeneous in real-world situations. In such situations, superpixels can be used to address non-homogeneous haze by segmenting an image in to a number of closely spaced patches. In this research, we develop an adaptive non-homogeneous image dehazing method for segmenting various hazy regions using superpixel-guided algorithm. As the estimation of both transmission map and ambient/atmospheric light made a substantial influence on the outcomes, with this viewpoint, the paper is focused in developing the model for image dehazing by considering non-uniform transmission map and non-uniform ambient/atmospheric light for different haze levels. Quantitative analysis demonstrates the effectiveness of our proposed model.