Enhanced Deep Dehazing for Haze Removal in License Plates

This paper introduces Enhanced Deep Dehazing (EDD), a novel deep learning model that overcomes the limitations of traditional image dehazing methods. Unlike approaches that rely on complex transmission map estimation, EDD utilizes an Encoder-Decoder architecture for direct data-driven learning. EDD effectively handles diverse hazy conditions that degrade image quality across various applications. It utilizes convolutional layers for feature extraction and deconvolutional layers for image reconstruction, employing ReLU activation and batch normalization to improve training stability and convergence. A case study on the application of EDD to car license plate dehazing, a critical task for autonomous vehicles and surveillance systems, is presented. EDD was trained and tested on the Foggy-Hazy Licence Plates dataset, with experimental results demonstrating superior efficiency compared to previous works, thus underscoring its effectiveness and potential in dehazing foggy/hazed license plates and highlighting its applicability in realworld scenarios.