数学形态学与幻想脸重建

Mathematical morphology and hallucinating face image reconstruction

提出了基于数学形态学的幻想脸重建算法,设计了表达图像特征的张量结构,介绍了幻想脸的重建步骤。算法首先将待重建图像的整体特征按区域分块形成特征模板,然后用这些特征模板在低频特征集中搜索特征距离最小的邻域,最后把该邻域对应的高频信息通过最优化的线性组合分块实现人脸图像的超分辨率重建。实验结果表明,该算法获得的图像视觉效果明显改善,峰值信噪比(PSNR)高于其它算法2.6dB以上。

A novel algorithm is proposed for hallucinating face reconstruction based on mathematical morphology.The algorithm designed a tensor structure to represent image features and shows the detailed process of hallucinating face reconstruction.The global features of an image is firstly divided into different feature templates according to their regions,then these feature templates are used to search neighbor-regions with the minimum feature distance in low frequency feature sets,and finally the super-resolution ...