Gunturk Bahadir K, Batur Aziz U, Altunbasak Yucel, Hayes Monson H, Mersereau Russell M
Sch. of Electr. and Comput. Eng., Georgia Inst. of Technol., Atlanta, GA 30332-0250, USA.
IEEE Trans Image Process. 2003;12(5):597-606. doi: 10.1109/TIP.2003.811513.
Face images that are captured by surveillance cameras usually have a very low resolution, which significantly limits the performance of face recognition systems. In the past, super-resolution techniques have been proposed to increase the resolution by combining information from multiple images. These techniques use super-resolution as a preprocessing step to obtain a high-resolution image that is later passed to a face recognition system. Considering that most state-of-the-art face recognition systems use an initial dimensionality reduction method, we propose to transfer the super-resolution reconstruction from pixel domain to a lower dimensional face space. Such an approach has the advantage of a significant decrease in the computational complexity of the super-resolution reconstruction. The reconstruction algorithm no longer tries to obtain a visually improved high-quality image, but instead constructs the information required by the recognition system directly in the low dimensional domain without any unnecessary overhead. In addition, we show that face-space super-resolution is more robust to registration errors and noise than pixel-domain super-resolution because of the addition of model-based constraints.
监控摄像头捕捉到的面部图像通常分辨率很低，这严重限制了人脸识别系统的性能。过去，已经提出了超分辨率技术，通过合并来自多幅图像的信息来提高分辨率。这些技术将超分辨率作为预处理步骤，以获得高分辨率图像，随后将其传递给人脸识别系统。考虑到大多数先进的人脸识别系统使用初始降维方法，我们建议将超分辨率重建从像素域转移到低维面部空间。这种方法具有显著降低超分辨率重建计算复杂度的优点。重建算法不再试图获得视觉上改善的高质量图像，而是直接在低维域中构建识别系统所需的信息，而没有任何不必要的开销。此外，我们表明，由于添加了基于模型的约束，面部空间超分辨率比像素域超分辨率对配准误差和噪声更具鲁棒性。
