In a previous blog post, Face Hallucination was introduced as a specialization of Image Super Resolution (SR) that also refers to Face Super Resolution (FSR). We’ve also listed the constraints postulated by Liu et al. for a successful FH. Let’s list them here once again for the sake of completeness:
1. Sanity constraint: the target HR image should be very close to the input LR image when smoothed and down-sampled.
2. Global constraint: the target HR image should have the common characteristics of human faces, e.g., possessing a mouth and a nose, being symmetrical, etc.
3. Local constraint: the target HR image should have the specific characteristics of the original LR face image, with photorealistic local features.
In this blog post, we propose a novel approach to FH based on Gu et al.  in which they argued that a Convolutional Sparse Coding (CSC) would achieve superior performance to Yang et al.  Sparse Coding Super Resolution (ScSR). Hence, we apply the CSC techniques in the context of Liu et al.  framework for FH.
As a side note, using Deep Convolutional Networks may surpass Sparse Representation approaches to Image Super Resolution. As suggested by .
1. S. Gu, W. Zuo, Q. Xie, D. Meng, X. Feng, L. Zhang, “Convolutional Sparse Coding for Image Super-resolution,” in ICCV 2015. [paper] [code]
2. Learning a Deep Convolutional Network for Image Super-Resolution
4. Ce Liu, Heung-Yeung Shum & William T. Freeman. Face Hallucination: Theory and Practice. International Journal of Computer Vision, Volume 75 Issue 1, October 2007, Pages 115-134.
5. Jianchao Yang, Hao Tang, Yi Ma & Huang, T. Face hallucination VIA sparse coding. Image Processing, 2008. ICIP 2008. 15th IEEE International Conference on 12-15 Oct. 2008, Page(s): 1264 – 1267.
6. Image Super-Resolution Using Deep Convolutional Networks
Chao Dong, Chen Change Loy, Kaiming He, Xiaoou Tang http://mmlab.ie.cuhk.edu.hk/projects/SRCNN.html