基于生成对抗网络的车牌图像超分辨率重建方法

A Super Resolution Reconstruction Method of License Plate Image Based on GAN

针对现有的车牌图像超分辨率重建方法网络复杂度较高的问题,提出一种轻量化的基于生成对抗网络的车牌图像超分辨率重建方法(S-ESRGAN)。借鉴ShuffleNet v2的设计概念,在生成器网络中引入了无序挤压和激励模块,在保证重建图像质量的前提下实现了轻量化。通过在CCPD数据集上进行测试,并与目前流行的算法对比,证明S-ESRGAN重建的车牌图像具有较高的PSNR和SSIM,计算量和重建时间方面明显优于重建图像效果最好的ESRGAN。

A lightweight super-resolution reconstruction method(S-ESRGAN)for license plate images based on generative adversarial networks is proposed to address the problem of high network complexity of existing methods for super-resolution reconstruction of license plate images.The unordered squeezing and excitation modules are introduced in the generator network with the design concept of ShuffleNet v2,to achieve light weighting while ensuring the quality of the reconstructed images.The test on the CCPD dataset and the comparison with the currently popular algorithmsshow that the license plate images reconstructed by S-ESRGAN have higher PSNR and SSIM,and are significantly better than ESRGAN,which has the best reconstruction image effect,in terms of computation and reconstruction time.