| Title |
Deep Learning Method for Extending Image Intensity Using Hybrid Log-Gamma |
| Authors |
(Yunju Kim) ; (Changwoo Lee) |
| DOI |
https://doi.org/10.5573/IEIESPC.2020.9.4.312 |
| Keywords |
Deep learning; Hybrid Log-Gamma; Extending image intensity; U-net; SSIM |
| Abstract |
The resolution, intensity range and color gamut of the latest display devices such as organic light-emitting diode (OLED) displays have been improved significantly compared to conventional display devices. Regarding the intensity range, the latest display devices can produce a maximum luminous intensity of 1,000 nits or more, which extends the intensity range considerably compared to conventional displays, which have a maximum luminous intensity of approximately 300 nits. In this paper, a deep learning-based image intensity range extension method is studied. The input and target images for deep learning are generated from the high dynamic range (HDR) images, and the target images have an extended intensity range using the Hybrid Log-Gamma (HLG) curve. A modified structure of U-net is proposed to improve the convergence of U-net, and an efficient learning method through the adoption of a structural similarity (SSIM) loss as a loss function is also proposed. The extensive simulations reveal the significantly improved performance of the proposed method. |