Analyzing image deblurring algorithms for ground-based sky cameras

• Published in: Remote sensing applications 2022-11, Vol.28, p.100813, Article 100813
• Main Authors: Jain, Navya, Jain, Mayank, Dev, Soumyabrata
• Format: Article
• Language: English
• Subjects: Deblurring; Deep Learning; Image evaluation; Image processing; Peak signal-to-noise ratio (PSNR); Structural Similarity Index (SSIM)
• ISSN: 2352-9385; 2352-9385
• DOI: 10.1016/j.rsase.2022.100813

Ground-based sky/cloud imagers are used widely by researchers to study various atmospheric activities. These imagers generate images in large quantities with high temporal and spatial resolution at a low cost. However, the images generated by these imagers are prone to noise and blur due to various reasons such as atmospheric dust, bird flocks, and glare from the sun. Removal of this blur becomes essential because the presence of blur in these images can cause a hindrance in the analysis of cloud images. Many algorithms have been proposed in the literature to deblur images. As far as we know, none of these algorithms have been applied to sky/clouds images. These algorithms are designed to remove blur from images with fixed edges in the original image. The effect of these algorithms can be different for sky/cloud images since clouds naturally have fuzzy borders and different shapes and textures. We apply a few popular algorithms to these cloud images after classifying them based on cloud class to examine their effects on cloud images in this paper. Then through an extensive image evaluation and in-depth analysis of all the algorithms, we try to find an apt algorithm for deblurring each cloud class. •We investigate the impact of deblurring algorithms on sky/cloud images.•We examine the effectiveness of such algorithms on images with fuzzy edges.•We analyze impact of state-of-the-art deblur methods on cloud type recognition.•We conclude that variance in sharpness plays a crucial role in deblurring results.