Among current true 3D display technologies, multi-layer 3D displays based on the principle of compressive light field have the advantages of high resolution, simple structure and faithful restoration of depth cues, demonstrating enormous research value and application potential. In recent years, multi-layer 3D displays have attracted increasing attention from researchers and some progresses have been made in improving the performance. However, there are still some limitations, such as the color deviation issue which causes unnatural colors of the reconstructed scene. In this paper, we propose using a customized look-up table (LUT) to alleviate the color deviation problem of multi-layer displays. For each of the display layers, we measured the response curves of the RGB channels, respectively, corresponding to different input gray levels. Then we compared them with a commercial standard display, so that we could correct each value within the gray range of the three channels to obtain a target output response, and the corrected values were used to build the look-up table. Using the customized LUT, we successfully achieved correction of color deviation in our multilayer display system. Finally, we demonstrated a 3D scene with natural colors, proving the effectiveness of our method on correcting the color deviation in multi-layer light field displays.
After decades of development, artificial neural network has become one of the most important research directions of artificial intelligence, and has a wide range of applications and important value in computer vision, natural language processing and other fields. Today, most of the applied artificial neural networks are based on von Neumann electronic hardware. As the semiconductor process approaches the physical limit, the performance growth encounters bottlenecks, and the power consumption problem is difficult to solve, limiting the application and further development of deep learning. Optical neural network provides a way to break through the bottleneck due to its high speed, high parallelism and low power consumption. At present, most optical neural networks are difficult to expand the depth of the network and have limited performance. In this paper, a multilayer optoelectronic hybrid convolutional neural network with an optical 4f-system recurrent structure is proposed. The electronic convolutional layer is replaced by an optical convolutional layer based on the 4f system, and the depth of the neural network is extended by the recurrent structure of the 4f system to improve its performance. Experiments show that the recognition accuracy of CIFAR-10 dataset of the proposed hybrid neural network is close to that of a corresponding electronic neural network. This work provides a possible way to build a deeper optoelectronic hybrid convolutional neural network when dealing with complicated problems.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.