The Low-Illumination Image Enhancement Methodof Aerial Photography Based on EnlightenGAN Network was Improved
In order to solve the problems of insufficient brightness,blurred details and noise in aerial images under low light conditions,an image enhancement method based on SC-Enlight-enGAN network was proposed.Firstly,low-light images and attention grayscale maps were used as inputs.Secondly,the self-regularized guided U-Net network with channel attention mecha-nism and position information attention mechanism was used as the generator to reduce noise and retain more image details in the enhancement process.Then,a global-local discriminator was used to ensure that the overall brightness of the image was consistent.Experimental results showed that the SC-EnlightenGAN network had an average increase of 1.565dB in the peak sig-nal-to-noise ratio and a 2％increase in the structural similarity index compared with the En-lightenGAN network.The improved model can effectively improve the brightness and subject detail information of aerial low-light images.
NETL
NSTL
万方数据
