Due to the absorption and scattering properties of water media，underwater images often suffer from imaging distortions，such as fogging，low contrast and color degradation，which seriously affect the subsequent utilization. To restore clear underwater images，an improved GAN-based deep learning model is proposed for underwater image enhancement. By employing the image quality assessment technique，the generated intermediate samples are fitted to the high quality samples，and the difference information obtained from the fit is used to optimize the generators in the network. The improved generative adversarial network effectively ameliorates the problem of image quality improvement limitations brought by the true-false training logic. The experimental results show that the proposed method can effectively restore the color of the underwater images，and improve the clarity and contrast. In addition，comparing with several state-of-the-art methods，our proposed method achieves increases of 2.9%，6.2% and 14.3% in terms of SSIM，UCIQE and UIQM indexes，respectively.
