Zero-Reference Image Restoration for Under-Display Camera of UAV

Abstract

The exposed cameras of UAV can shake, shift, or even malfunction under the influence of harsh weather, while the add-on devices (Dupont lines) are very vulnerable to damage. We can place a lowcost T-OLED overlay around the camera to protect it, but this would also introduce image degradation issues. In particular, the temperature variations in the atmosphere can create mist that adsorbs to the T-OLED, which can cause secondary disasters (i.e., more severe image degradation) during the UAV's filming process. To solve the image degradation problem caused by overlaying T-OLEDs, in this paper we propose a new method to enhance the visual experience by enhancing the texture and color of images. Specifically, our method trains a lightweight network to estimate a low-rank affine grid on the input image, and then utilizes the grid to enhance the input image at block granularity. The advantages of our method are that no reference image is required and the loss function is developed from visual experience. In addition, our model can perform high-quality recovery of images of arbitrary resolution in real time. In the end, the limitations of our model and the collected datasets (including the daytime and nighttime scenes) are discussed.

UDC Image Restoration via Lightweight Dual-Domain Network

“…They 16,17 also incorporate the domain knowledge into the model, which allows better recovery of the synthetic UDC images but may limit the application scenarios and increase the complexity of the network. Besides considering scenarios on the mobile phone device, there is also some current work that considers deconstructing a degraded UDC image on a UAV device 8,9 . Unlike the above work, which relies on a huge model capacity to recover images, we attempt to develop a lightweight model to efficiently reconstruct UDC images in this paper.…”

“…Secondly, some researchers introduce modules commonly used in other tasks 4,6 , without considering the degradations specific to UDC images in spatial and frequency domains, such as low lighting and spectrum attenuation. Thirdly, some methods incorporate the complex domain knowledge of UDC into the models 4,7 , which makes the model difficult to migrate to new UDC scenarios 8,9 . In this paper, we propose a lightweight dual-domain network for restoring UDC images (LD 2 Net) to integrate efficiency, accuracy, and interpretability.…”