基于超深残差通道注意力网络的图像压缩感知重构

Image Compressive Sensing Reconstruction Using Ultra-Deep Residual Channel Attention Network

提出了一种基于超深残差通道注意力网络的图像压缩感知重构算法。超深残差通道注意力网络的重构部分由多个残差组组成,每个残差组包含一个长连接和一组带有短连接的残差块。长连接结构能够有效传播丰富的低频信息,使主干网络专注于学习高频信息。在残差块中引入通道注意力机制,通过考虑通道间的相互依赖性来自适应地重新划分通道特征,从而达到强化重要特征的效果。实验表明,该算法能够有效提升图像压缩感知的重构精度。

An image compressive sensing reconstruction method based on ultra-deep residual channel attention networks is proposed.The reconstruction part of the ultra-deep residual channel attention network consists of multiple residual groups,each of which contains a long connection and a set of residual blocks with short connections.The long-connected structure can effectively deliver rich low-frequency information,allowing main network to focus on learning high-frequency information.The channel attention mechanism is introduced into residual blocks.By considering the interdependence between channels,the channel features keep changing adaptively so as to strengthen the important features.Experiments show that this method can effectively improve the reconstruction accuracy of image compressed sensing.