Two-field sequential color liquid crystal displays with deep learning-enabled real-time driving

Abstract

Two-field driving is the ultimate goal of field sequential color liquid crystal displays (FSC-LCDs) because it requires the lowest refresh rate and transmission bandwidth in addition to the intrinsic advantages of FSC-LCDs, e.g., tripled light efficiency and spatial resolution. However, fewer fields create a more significant challenge in controlling color breakup and distortion, as well as higher computational complexity in calculating LC signals. Regarding the difficulties, we propose a two-field FSC driving method that synchronously generates backlight and LC signals through two lightweight neural networks. The runtimes of the two networks are as fast as 1.23 and 1.79 ms per frame on a GeForce RTX 3090Ti graphic card, fully supporting real-time driving. Next, an over-partitioning approach is proposed to overcome the cross talk between backlight segments while processing high-resolution images. Besides the real-time feature, a reduction of 14.88% in color breakup concerning current methods and low distortion are verified. We also provide our open-source code.