Real-Time Defogging using Dark Channel Prior for Enhanced Visibility

The presence of fog is a notable obstacle for capturing images, particularly in the early morning hours. The diminished sight impedes the capacity to get distinct and comprehensive information from the surroundings. In order to tackle this issue, a rapid and live vision-based dehazing method is suggested. The objective of the technology is to dissipate the fog and enhance visual acuity. It does this by implementing four steps: using an exponential moving average filter to diminish the impact of noise and enhance clarity, preprocessing the filtered frames, detecting fog with the Gaussian blur technique, and applying the dark channel prior approach to achieve optimum contrast and remove fog. Comprehensive testing were carried out on live recordings to showcase the system's ability to significantly improve sight in different foggy situations. This technique demonstrates the superior performance in terms of following parameters like, an accuracy, precision, recall, and F1 score compared to previous investigations. Our methodology achieved the greatest accuracy when compared to state-of-the-art transfer learning convolutional neural networks.