P‐8.2: CNN for Dual‐layer LCDs

Abstract

Low contrast ratio (CR) is one big drawback for LCD, which is caused by light leakage of LC cell. Light leakage can be greatly reduced by bonding two LC panels to form a dual‐layer LCD. However, with the certain distance between the two LC panels, the ghost may appear when viewing off‐axis. This paper proposes a convolutional neural network‐based method to improve the display quality of a dual‐layer LCD. The network consists of 8 convolution layers, in which the kernel of the layer 4 and layer 8 is 1×1 to eliminate the redundant features and accelerate the processing speed. The kernel of the other layers is 3×3. Concat1 and Concat2 are included to fuse the features from the convolution layer 1‐3 and 5‐7. The loss function is especially defined with the consideration of the total differences between the reconstructed images at the different viewing angles and the input image. The simulation results show that compared with the existing methods, the proposed method has the merits of higher display quality and shorter processing time, which presents high practicability.