基于特征空间的人脸超分辨率重构

Feature Space Based Face Super-resolution Reconstruction

超分辨率图像重构是利用关于同一场景的多帧低分辨率图像重构出一幅具有更高分辨率图像的过程.传统的超分辨率图像重构算法是基于像素空间,通过利用高、低分辨率像素空间之间的映射关系来求解,具有计算复杂性高等缺点.针对低分辨率人脸放大问题,提出了一个基于特征空间的人脸超分辨率图像重构算法.与传统算法相比,该算法不仅降低了计算复杂性,还具有更好的鲁棒性.

Super-resolution image reconstruction is a process of producing a high-resolution image from a set of lowresolution images of the same scene. Previously published techniques usually perform reconstruction on pixel domain, which utilize the relationship between high-resolution and low-resolution pixel space, and suffer a high computational complexity. A feature space based super-resolution face reconstruction algorithm is proposed in this paper for application of high-resolution face reconstruction from low-resolution video surveillance. We showed that compared with normal algorithms, the proposed algorithm is more robust and more computationally efficient