基于改进残差网络的运动目标模糊图像复原方法

Moving object blurred image restoration method based on improved residual networks

传统的残差网络在复原运动目标模糊图像时,在模糊程度较严重的情况下,存在特征提取不充分、噪声干扰等问题,导致恢复出的图像无法完全达到原始图像的清晰度和细节。对此,提出基于改进残差网络的运动目标模糊图像复原方法。对采集到的运动目标模糊图像,采用多损失函数融合方法改进传统残差块结构,构建编码器-解码器网络训练结构,训练损失函数,提升网络的特征学习能力。通过完成训练的网络,输出运动目标模糊图像复原结果。实验结果表明,该方法复原运动目标模糊图像的峰值信噪比高于30 dB,结构相似性高于0.9。

When restoring blurred images of the moving objects,the traditional residual networks often suffer from insufficient feature extraction and noise interference in the case of serious blur occurs,so that the restored images fail to fully achieve the sharpness and details of the original images.Therefore,an improved residual network based method for restoring blurred images of moving objects is proposed.For the collected blurred images of moving objects,a multi loss function fusion method is adopted to improve the traditional residual block structure,construct an encoder-decoder network training structure,train the loss function,and enhance the feature learning ability of the networks.By the trained network,the restoration results of blurred images of the moving objects are output.The experimental results show that the peak signal-to-noise ratio(PSNR)of the restored blurred images of the moving objects is higher than 30 dB,and the image structural similarity is higher than 0.9.