Abstract: A new learning-based super-resolution algorithm is presented. Pyramid is used to extract the facial gradient distribution features, the standard face training database is established for the study model, these features are combined with pyramid-like parent structure to predict the best prior. And through the Bayesian Maximum A Posterior(MAP) frame, the high resolution face image is captured. Experimental results show that the proposed algorithm synthesizes high-resolution faces and eliminates the noise with better visual effect, and the average of peak signal-to-noise ratios is improved about 1.19 dB to 2.4 dB compared with some existing face super-resolution algorithms.

Key words: super-resolution, Bayesian, Maximum A Posterior(MAP), pyramid, parent structure

摘要： 提出一种基于学习的金字塔人脸超分辨率算法，利用金字塔学习人脸图像梯度的空间分布特性，建立标准人脸训练库作为学习模型，采用塔状父结构从训练库搜索匹配特征信息相似度最高的小块，预测出最优的拉普拉斯金字塔先验模型，利用贝叶斯MAP框架求出高分辨率人脸图像。实验结果表明，与其他人脸超分辨率算法相比，在将人脸图像分辨率提高4×4倍的情况下，该算法生成的高分辨率人脸图像的平均峰值信噪比提高1.19 dB~2.4 dB，可以更好地消除噪声，具有较好的视觉效果。

关键词: 超分辨率, 贝叶斯, 最大后验概率, 金字塔, 父结构

CLC Number: 

• TP18