Enhancing Computer Generated Hologram Quality via Deep Learning Based On U-Net Architecture
No Thumbnail Available
Date
2024-05-30
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This thesis investigates the advancement of computer-generated hologram (CGH) quality using
deep learning methodologies, with a focus on the U-Net architecture. Employing the DIV2k
dataset as the foundation, the study proposes a novel approach integrating a specialized 3U
configuration within the U-Net framework. This configuration comprises meticulously designed
U-shaped modules tailored to extract intricate details and subtle nuances inherent in holographic
imagery. Evaluation metrics such as Peak Signal-to-Noise Ratio (PSNR), Structural Similarity
Index (SSIM), and reconstruction time are employed to assess the model's performance.
Comparative analysis against established models, namely Holonet and Holoencoder, underscores
the superior efficacy of the proposed method in faithfully reproducing holographic images. The
findings reveal substantial enhancements in hologram quality, characterized by elevated PSNR
and SSIM values, alongside notable gains in computational efficiency manifested through reduced
reconstruction times. This research contributes significantly to the progression of holographic
imaging technology, offering promising avenues for future exploration and innovation in the realm
of computer-generated holography.