Underwater Image Enhancement Technology Based on Deblur GAN

Due to the instability of underwater environment, underwater images may have degradation phenomena such as color deviation, low color contrast and motion blur. To solve these problems, this paper proposes an enhancement algorithm for underwater images, which needs to go through two stages of color restoration and deblurring. In the first stage, gaussian filtering and mean shift are used to sharpen the image. Then，the image color is corrected by comparing the mean value of each color channel. Finally, the contrast of the image is adjusted by linear stretch. In the second stage, the Generative Adversarial Networks (GAN) with residual are used to extract the features of the image. With nine continuous residual networks, blurring effect is eliminated and image features are enhanced. When the proposed algorithm is used to process underwater images, it is found that it cannot only remove the image blur, but also eliminate the color deviation of the image without carrying red artifacts. Also, by comparing the two indexes of Underwater Image Quality Measures (UIQM) and Underwater Color Image Quality Evaluation (UCIQE), it is found that the proposed algorithm has a better image processing effect.