Cascading Dense Network for Face Image Super-Resolution

Recently, Convolutional neural network (CNN) based methods have achieved great success in the field of face image Super Resolution (SR). However, most of these methods tend to produce over-smoothed outputs, and suffer from degradation when dealing with very Low Resolution (LR) face images. To solve these problems, this paper proposes an accurate CNN architecture based on contourlet transform, which can ultra-resolve a very low resolution face image of 16 × 16 or smaller pixel size to its larger version. First of all, we present multi-scale fusion CNN (MSFC) to fully detect and exploit features from LR inputs. And then, we formulate the SR problem as the prediction of contourlet transform coefficients, which is able to make MSFC further capture the texture details for super-resolve face images. Extensive experiments demonstrate that the proposed method achieves superior face SR results and outperforms the state-of-the-art.