RBPNET: An Asymptotic Residual Back-Projection Network for Super Resolution of Very Low Resolution Face Image

The super resolution of a very low resolution face image is a challenge task in computer vision, because it is difficult to learn a non-linear mapping of input-to-target space by deep neural network in one step upsampling. In this paper, we propose an asymptotic Residual Back-Projection Network (RBPNet) to gradually learn residual between the reconstructed face image and the ground truth by self-supervision mechanism. We map the reconstructed high-resolution feature map back to the original low-resolution feature space, use the original low-resolution feature map as a reference to self-supervising the learning of the various layers. The real high-resolution feature maps are approached gradually by iterative residual learning. Meanwhile, we explicitly reconstruct the edge map of face image and embed it into the reconstruction of high-resolution face image to reduce distortion of super-resolution results. Extensive experiments demonstrate the effectiveness and advantages of our proposed RBPNet qualitatively and quantitatively.