GamutNet: Restoring Wide-gamut Colors for Camera-captured Images, a study on image color gamut restoration in which MAXST's Senior Researcher Tae-Hong Jeong participated as the first co-author, was presented at the International Conference CIC held by IS&T!đ
CIC is an annual technical gathering for scientists, technicians and engineers working in the field of color imaging.
The research conducted by Tae-Hong Jung, Principal Researcher at MAXST with a Ph.D. at Ajou University, Hyun-Jun Shin, Professor in the Department of Media at Ajou University, Professor Michael S. Brown in the Department of EECS at York University, Abdelrahman Abdelhamed at the Samsung AI Center in Toronto, and Hoang Le, a Ph.D. Researcher in the Department of EECS at York University is about a deep neural network that reconstructs standard gamut* images taken by a camera into wide gamut images.
Color gamut* refers to the range of colors included in a digital image or the range of colors that a display can express. In general, the wider the color gamut of digital images and displays, the more vivid colors can be expressed!
As a result of an experiment in which a new data set was built based on the research team's in-camera processing pipeline, the deep neural network trained using it showed superior color gamut restoration performance compared to the existing method.
On the image aboveâ, the Input sRGB image in the first column is the original, and the Ground truth image in the sixth column is the restoration target image.
In addition, the images in the second to fourth columns are the contents restored through the existing methods such as Zamir et al., Photoshop, and Color Sync, respectively. And the images in the fifth column are the results of restoration using GamuNet, a new method of this study.
Also, the images in the second and third rows are about visual error and color distribution within the color gamut, respectively.
It can be seen that the restoration results using Zamir, Photoshop, and Color Sync have a large visual error and color distribution compared to the target image. On the other hand, it can be seen that GamuNet produced a result close to the target image compared to the others.
With the development of technology, the demand for wide color gamut content is increasing, where high-performance display panels are distributed at low prices.đ
It is getting very cold these days, so we wish you all take good care of yourself!
If you are curious about MAXST's Virtual Convergence Technology (XR), please check the MAXST platform and solutionâs official websites below.
MAXST Homepage: https://maxst.com/
MAXST AR SDK Developer Site: https://developer.maxst.com/
MAXWORK Homepage: https://maxwork.maxst.com/
MAXST VPS Homepage: https://vps.maxst.com/
Comments