Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AutoColor: Learned Light Power Control for Multi-Color Holograms

Yicheng Zhan, Koray Kavaklı, Hakan Urey, Qi Sun, Kaan Akşit

TL;DR

Multi-color holography promises higher dynamic range but requires efficient control of light-source powers. AutoColor is a learned network that, from input images, predicts the $3\times3$ light-source power matrix $l$, using a permutation-invariant loss to handle frame ordering; it reduces optimization from >1000 steps to about 70 steps while preserving reconstruction quality, validated on a hardware prototype with real lasers and an SLM. The dataset is assembled via GPT-4 guided prompts, Stable Diffusion, Real-ESRGAN, and monocular depth estimation, and ground-truth powers come from prior co-optimization, enabling fast, interactive-rate multi-color holography. This work advances practical high-dynamic-range holographic displays and sets the stage for extensions to varied brightness scales and gaze-driven color prioritization to save energy.

Abstract

Multi-color holograms rely on simultaneous illumination from multiple light sources. These multi-color holograms could utilize light sources better than conventional single-color holograms and can improve the dynamic range of holographic displays. In this letter, we introduce AutoColor , the first learned method for estimating the optimal light source powers required for illuminating multi-color holograms. For this purpose, we establish the first multi-color hologram dataset using synthetic images and their depth information. We generate these synthetic images using a trending pipeline combining generative, large language, and monocular depth estimation models. Finally, we train our learned model using our dataset and experimentally demonstrate that AutoColor significantly decreases the number of steps required to optimize multi-color holograms from > 1000 to 70 iteration steps without compromising image quality.

AutoColor: Learned Light Power Control for Multi-Color Holograms

TL;DR

Multi-color holography promises higher dynamic range but requires efficient control of light-source powers. AutoColor is a learned network that, from input images, predicts the light-source power matrix , using a permutation-invariant loss to handle frame ordering; it reduces optimization from >1000 steps to about 70 steps while preserving reconstruction quality, validated on a hardware prototype with real lasers and an SLM. The dataset is assembled via GPT-4 guided prompts, Stable Diffusion, Real-ESRGAN, and monocular depth estimation, and ground-truth powers come from prior co-optimization, enabling fast, interactive-rate multi-color holography. This work advances practical high-dynamic-range holographic displays and sets the stage for extensions to varied brightness scales and gaze-driven color prioritization to save energy.

Abstract

Multi-color holograms rely on simultaneous illumination from multiple light sources. These multi-color holograms could utilize light sources better than conventional single-color holograms and can improve the dynamic range of holographic displays. In this letter, we introduce AutoColor , the first learned method for estimating the optimal light source powers required for illuminating multi-color holograms. For this purpose, we establish the first multi-color hologram dataset using synthetic images and their depth information. We generate these synthetic images using a trending pipeline combining generative, large language, and monocular depth estimation models. Finally, we train our learned model using our dataset and experimentally demonstrate that AutoColor significantly decreases the number of steps required to optimize multi-color holograms from > 1000 to 70 iteration steps without compromising image quality.
Paper Structure (5 sections, 4 equations, 2 figures)

This paper contains 5 sections, 4 equations, 2 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Evaluation
  5. Conclusion

Figures (2)

  • Figure 1: AutoColor light power estimation network structure. AutoColor learns to estimate powers for each light source to illuminate multi-color holograms using our multi-color hologram dataset and a permutation-invariant loss tailored for multi-color holograms.
  • Figure 2: Photographs showing AutoColor generating x1.8 times brighter images in lesser steps than Multicolor. Experimental results show that AutoColor achieves high-fidelity visuals using only 70 steps, whereas Multicolor requires 1000 steps for similar quality and fails to produce correct color information in 70 steps. Images optimized with 70 steps using AutoColor provides similar quantitative image metrics (see the inset numbers) compared with the images generated with 1000 steps in Multicolor. AutoColor applies to both 2D images (first and second row) and 3D images (third row). (Source link: https://github.com/complight/images, 80 ms exposure).