基于CCA空间的平滑稀疏超分辨率人脸重构

摘要

直接基于人脸图像空间构建的高低分辨率字典无法满足高度相关的条件,影响了重构的效果.提出了一种基于典型相关分析(CCA)空间的平滑稀疏超分辨率人脸重构方法.将映射到CCA空间的字典进行排序,并进行稀疏更新;将得到的新字典根据输入测试块重新映射到CCA空间;引入平滑稀疏模型.实验结果表明:相对于其他人脸重构方法,所提方法能够取得更好的去噪效果,更清晰的重构效果以及良好的平滑性.%The constructed high and low resolution dictionaries based on face image space can not satisfy the highly relevant conditions,which affects the effect of these reconstruction methods. A smooth sparse super-resolution face reconstruction method based on canonical correlation analysis(CCA)space is proposed. Dictionaries mapped to the CCA space are sorted and sparse is updated,and the resulting new dictionary According to the input test block remapped to the CCA space,and introduce the smooth sparse model. The experimental results show that our method can achieve better denoising effect,clearer reconstruction effect and good smoothness than other face reconstruction methods.