Abstract
In the past decade, significant research effort has been directed toward developing single-image dehazing algorithms. Despite this effort, dehazing continues to present a challenge, particularly in complex real-world cases. Indeed, it is an ill-posed problem because scene transmission depends on unknown and nonhomogeneous depth information. This paper proposes a novel end-to-end adaptive enhancement dehazing network (AED-Net) to recover clean scenes from hazy images. We evaluate it quantitatively and qualitatively against several state-of-the-art methods on three commonly used dehazing benchmark datasets as well as hazy real-world images. Moreover, we evaluated it against the top-scoring methods of the Codalab NTIRE 2021 competition based on the dehazing challenge dataset. Extensive computer simulations demonstrated that AED-Net outperforms state-of-the-art single-image haze removal algorithms in terms of PSNR, SSIM, and other key metrics. Furthermore, it improves image texture, detail edges, boosts image contrast and color fidelity. Finally, AED-Net is more effective under complex real-world conditions.
Highlights
Images captured in hazy or foggy weather conditions can be significantly degraded, rendering the image objects and their features challenging to be identified by imaging systems
Liu et al [26] surveyed algorithms based on the physical model [1]–[3], image enhancement [17], and deep learning (DL) [7]–[16]
We propose a novel end-to-end adaptive enhancement dehazing network called AED-Net for single-image dehazing
Summary
Images captured in hazy or foggy weather conditions can be significantly degraded, rendering the image objects and their features challenging to be identified by imaging systems. Liu et al [1] demonstrated that the reduction in detection rate was proportional to the haze density, positing that image dehazing could be a practical solution for image classification, object detection, remote sensing, image or video retrieval, and outdoor surveillance [1]–[4], [50]. Several dehazing algorithms have been developed that rely on additional information, such as depth, polarization, and multiple images [4]–[6], as well as a single image [7]–[12]. Because multiple images or additional physical information are often unavailable, single-image dehazing has received the most attention. Liu et al [26] surveyed algorithms based on the physical model [1]–[3], image enhancement [17], and deep learning (DL) [7]–[16]
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
