基于多尺度融合卷积神经网络的图像去雾算法

Image dehazing algorithm based on multi-scale concat convolutional neural network

为解决图像去雾后颜色偏暗以及去雾不彻底等问题,本文提出了一种基于多尺度融合卷积神经网络的图像去雾算法。以有雾图像为输入,首先经过预处理模块由单尺度卷积层提取有雾图像浅层信息,然后设计多尺度映射模块实现深度特征学习以及深、浅层特征融合,由反卷积模块还原图像尺寸,通过卷积操作得到有雾图像对应的粗透射率图。采用双边滤波法优化输出细透射率图,最后依据大气散射模型复原出无雾图像。实验结果表明本文方法在合成有雾图像和自然有雾图像上均优于其他算法,其中合成有雾图像上的峰值信噪比(PSNR)、结构相似性(SSIM)能分别达到29.238、0.950。本文所提算法可以有效地避免去雾图像颜色偏暗、失真等不足,提高了图像去雾性能并体现出良好的视觉效果。

In order to solve the problem of dark color and incomplete defogging after image defogging,an image defogging algorithm based on multi-scale concat convolutional neural network is proposed in this paper.Taking the foggy image as the input,the shallow layer information of the image is extracted from the single scale convolution layer through the preprocessing module,and then the multi-scale mapping module is designed to realize the depth feature learning and the fusion of the deep and shallow layer features.The deconvolution module is used to restore the image size,and the coarse transmittance map corresponding to the foggy image is obtained through the convolution operation.Finally,the haze free image is restored according to the atmospheric scattering model.The experimental results show that the proposed method is superior to other algorithms in both synthetic and natural foggy images,and the peak signal-to-noise ratio(PSNR)and structure similarity(SSIM)can reach 29.238 and 0.950,respectively.The proposed algorithm can effectively avoid the dark color and distortion of the image,improve the image defogging performance and show good visual effect.