A Transmittance Optimization-based Framework for Image Dehazing on Multi-rotor Drones Imaging

The imaging quality of images collected by multi-rotor drones determines its practical application effects. However, current image enhancement dehazing methods have problems such as being affected by depth information, high computational complexity, and artefacts in the restoration results. In this paper, based on the dark channel prior model, a tolerance mechanism is introduced in the transmittance estimation part. The total variation (TV) model constrained by the ℓ1-norm is used to refine the transmittance estimation. In addition, to reduce our algorithm’s calculation, we use down-sampling technology to reduce the original image to obtain the transmittance part. Then we calculate the transmittance of the small resolution image. In the end, we can generate the transmittance of the original image by interpolation. The experimental data processing results verify the effectiveness of our algorithm.