视觉盲计算技术研究进展

 云计算和大数据的广泛应用使得多媒体数据的隐私保护面临越来越严峻的挑战，而多方计算、同态密码、函数加密等密码技术为数据在密文下的盲计算奠定了基础。视觉盲计算是在不接触图像、视频等视觉数据原始内容的情况下对其进行检测、识别、检索及更复杂的处理，是计算机视觉与密码学等领域学科交叉的新方向，在视频监控、多媒体数据共享、云计算、移动计算等领域有广泛的应用前景。本文回顾视觉盲计算技术的发展历史，从隐私保护的视频监控、人脸检测、人脸识别、人脸检索、机器学习等方面综述了该领域关键技术的发展情况，展望了视觉盲计算技术的发展趋势。

Recent advances on blind vision

With the rapid development of multimedia technology and the Internet, privacy protection of multimedia data is facing more and more challenges. The widespread application of cloud computing and big data, on the other hand, has further exacerbated people's concerns about privacy security. However, multiparty computation, homomorphic cryptography, functional encryption and other cryptographic techniques have made possible the blind computation of data under ciphertext. In particular, visual blind computing is to detect, identify, search, and deal with data without touching their original contents such as images and videos. As a new direction in the combination of computer vision and cryptography, it has a wide range of applications in video monitor, multimedia data sharing, cloud computing, mobile computing and other fields. This paper reviews the development history of visual blind computing technology, summaries the key technologies for video monitor, face detection, face recognition, face retrieval, machine learning and so on, and looks into the development trend of visual blind computing technology.