![]() ![]() Semantic face hallucination: Super-resolving very low-resolution face images with supplementary attributes. IEEE Transactions on Image Processing, 30, pp.1728-1743. ![]() Face hallucination with finishing touches. Zhang, Y., Tsang, I.W., Li, J., Liu, P., Lu, X. Supervised Pixel-Wise GAN for Face Super-Resolution. IEEE Transactions on Image Processing, 30, pp.1219-1231. Learning Spatial Attention for Face Super-Resolution. Machine Vision and Applications, 31 (4), pp.1-12.Ĭhen, C., Gong, D., Wang, H., Li, Z. Improved face super-resolution generative adversarial networks. International Journal of Computer Vision, 128 (2), pp.500-526. Hallucinating unaligned face images by multiscale transformative discriminative networks. ![]() ID Preserving Face Super-Resolution Generative Adversarial Networks. IEEE Transactions on Image Processing, 28 (12), pp.6225-6236. Sigan: Siamese generative adversarial network for identity-preserving face hallucination. Verification of very low-resolution faces using an identity-preserving deep face super-resolution network. Springer, Cham.Ītaer-Cansizoglu, E., Jones, M., Zhang, Z. In Pacific Rim Conference on Multimedia (pp. Enhanced discriminative generative adversarial network for face super-resolution. Yang, X., Lu, T., Wang, J., Zhang, Y., Wu, Y., Wang, Z. Pattern Recognition Letters, 111, pp.72-79. High-quality face image generated with conditional boundary equilibrium generative adversarial networks. Due to the observation that the degradation process in the real world is too complex to be simulated. Unsupervised or self-supervised methods will become mainstream face super-resolution methods. Existing face super-resolution methods mainly focus on the case of the magnification factors x8. More challenging scales, such as x32, 圆4, can be explored. Hence, developing models with more lightweight and lower computation cost is still a major challenge. Lightweight Face Super-resolution Models:ĭeep learning-based face super-resolution methods have achieved great breakthroughs, they have difficulty in deploying real-world applications, which is caused by a mass of parameters and high computation cost. However, the balance between subjective and objective quality is important and existing face super-resolution methods ignore how to find a balance between them. Thus, higher-dimensional prior can significantly enhance face reconstruction.īalance between Subjective and Objective Quality:įace super-resolution tends to recover SR with higher PSNR but poorer visual quality and vice versa. High-dimensional Facial Prior Information:įace super-resolution methods is increasingly complex and has higher than higher dimensions, from 2D images (facial landmarks, facial heatmaps, parsing maps) to 3D prior, which means higher-dimensional prior provides richer information. ![]()
0 Comments
Leave a Reply. |