A generic post-processing framework for image dehazing

There are several methods available for image dehazing. The challenges faced by most of these algorithms include under-exposure and leftover haze after dehazing, which eventually leads to low brightness and low contrast, respectively. Therefore, to overcome these drawbacks a post-processing method is required for image dehazing. Some post-processing techniques are implemented earlier for image dehazing such as contrast-limited adaptive histogram equalization, exposure enhancement, and adaptive tone remapping. However, these algorithms may not be applicable to all the dehazing methods as they may produce over-exposed, over-enhanced, and under-enhanced results. Hence, a generic post-processing (GPP) model is needed that can be applied to any dehazing algorithm and also overcome the drawbacks of the previous dehazing and post-processing techniques. A GPP framework is proposed in this paper which adaptively enhances the dehazed image based on its quality. The image quality assessment parameters called normalized brightness difference and haze score are introduced in this paper to detect the under-exposure and leftover haze in the dehazed images, respectively. If a dehazed image exhibits under-exposure, its brightness has to be improved. The fast low-light image enhancement is proposed to enhance the brightness of the under-exposed image. The contrast enhancement algorithm has to be applied when a hazy image exhibits leftover haze. The introduction of the proposed post-processing model adaptively enhances the dehazed image and also brings significant improvements for any dehazing method in terms of quantitative and qualitative assessments.