Deep CNN-based local dimming technologys

Local dimming is a feature on light-emitting diode (LED) TVs that dims the backlight behind parts of the screen that are displaying black and brights the backlight behind parts of the screen that are displaying white. To improve contrast ratios (CRs) and reduce power consumption for liquid crystal displays (LCDs), local dimming technology has been widely studied. In this paper, we propose a deep CNN-based local dimming technology for LCD-LED dual-modulation display, which includes backlight luminance extraction network (BLEN) and pixel compensation network (PCN). BLEN aims to generate backlight luminance while PCN aims to generate pixel-compensated images. To achieve this technology, we first use seven kinds of traditional backlight luminance extraction methods and five kinds of traditional pixel compensation methods to generate the pleasant backlight luminance and pixel compensated images as ground-truth. And then the proposed networks are trained and verified on the self-constructed image dataset. The experimental results show that the proposed deep CNN-based local dimming technology can improve the CRs and reserve more details than other methods, which further vertify on LCD-LED dual modulation display.