Desmogging of Smog Affected Images Using Illumination Channel Prior

Abstract

Visibility restoration of smoggy images plays a significant role in various computer vision applications. However, designing an efficient desmogging technique is still a challenging issue. The majority of existing researchers have designed restoration model for rainy, dusty, foggy, hazy, etc., images only. Therefore, these approaches perform poorly for smoggy images. In this paper, a novel illumination channel prior is proposed to restore smoggy images in a significant way. A gradient magnitude-based filter is also utilized to refine the transmission map. The proposed desmogging approach is compared with the existing visibility restoration approaches over ten real-time smoggy images. The subjective and quantitative analysis reveals that the proposed desmogging approach outperforms others.