Методы коррекции изображений, полученных в условиях пониженной видимости

Download article in PDF format

Authors: Elizarov A. I., Shaleev A. V.

Annotation: Image restoration in low-visibility conditions is an important issue. Numerous methods have been proposed to address it. However, there is no universal solution equally effective for all shooting conditions. In this work, we compare CLAHE, Reti-nex, the Dark Channel Prior (DCP), and the deep learning model DehazeNet. The BRISQUE no-reference metric was used to evaluate the quality of restored images. Tests on real-world data show that the Dark Channel Prior provides the best metric values among the approaches considered, confirming its versatility and reliability for practical computer vision applica-tions in fog and haze.

Keywords: deep learning, image correction, computer vision, fog