Hitoshi Kiya Laboratory
Professor Hitoshi Kiya
Signal Processing / Image, Video and Audio Processing / Data Compression / Social Media / Multimedia Security and Forensics
With the spread of the Internet and social media, it is indispensable in our daily life and social systems to utilize information expressed as images, video and audio. At the same time, problems with security and privacy, crimes and criminal investigations are closely related to these media. In this laboratory, we have been conducting media research from this background. Typical research subjects are shown below.
High quality image / video generation
As represented by keywords such as 8K and HDR (High Dynamic Range), the quality of images and video has been improved. At the same time, it is an urgent issue to improve the quality of images taken under harsh conditions in surveillance cameras and automatic driving of automobiles. From such a background, the field of “Computational Photography”, in which images are generated digitally, rather than simply taking an image of the subject optically, is receiving a lot of attention. Our laboratory has several research themes in this field.
Media security forensics
Digital media generally contain a lot of personal and copyright information. Therefore, digital media have caused many problems in the spread of social network services and cloud computing. In addition, the media can be easily modified and processed, so tamper detection and authenticity verification methods are important research subjects. Against this background, we are conducting research on computation, information retrieval, and computer forensics (scientific investigation) on encrypted data.
Machine learning / deep learning and media information processing
Traditionally, images and video have been generated for human visual recognition. However, with the expansion of their use, the field of computer vision for recognizing data with robots and computers has been rapidly developing. Typical examples include automatic driving, robot vision and image search. In these fields, deep learning, which is one of the machine learning methods, has given top data and is innovating the conventional technology. In our laboratory, we have been investigating what high-quality images for robots and computers are, and new possibilities for media information processing under the use of machine learning and deep learning.
