Research on Dehazing Algorithm of Single UAV Reconnaissance Image under Different Landforms Based on Retinex

Abstract

Aiming at the weakness of the dark channel prior modelin and Retinex algorithm such as slow calculation speed, halo effect, and loss of details in the processing of UAV single image dehazingand, an improved Retinex algorithm is proposed. First, convert the RGB color space to the HIS color space, separate the brightness I, assign values to the three scales of large, medium and small, use Gaussian surround to obtain the three scale brightness images and then weight, and then switch the brightness image to the RGB color space, in order to achieve the information with more details. Finally, the haze images which is distinguished in three different landforms such as city, mountainous area and forestwill be performed simulation experiments in MATLAB, and adjust the image restoration effect by changing the parameters.