基于CA-Res2Net和可变形卷积的图像去模糊方法

Image Deblurring Method Based on CA-Res2Net and Deformable Convolution

针对动态场景下的非均匀运动模糊问题,提出了一种基于细粒度多尺度注意力残差模块(CA-Res2Net)和可变形卷积(Deformable Convolution)的图像去模糊方法CADNet。CADNet以可变形卷积自编码结构为基础架构,进而可以更好地对抗模糊形变;同时CADNet嵌入了细粒度多尺度的残差模块,可以在更细粒度的层次上表示图片的多尺度特征信息,从而可以获得更强的特征表示能力。实验结果表明,CADNet是一种高效的去模糊算法,在获得更高去模糊性能的同时,压制住了计算开销。

Aiming at the problem of non-uniform motion blur in dynamic scenes,an image deblurring method CADNet is proposed,in which the main components are deformable convolution layer and fine-grained multi-scale attention residual module.The designation of deformable convolution layer can make CADNet be better resist fuzzy deformation,while the fine-grained multi-scale residual module can represent the features of images in a finer-grained level,and then improve the representation ability of CADNet.The experimental results demonstrate that CADNet is an efficient deblurring algorithm,which suppresses the computational overhead,while achieving higher deblurring performance.