Achromatic Extended-Depth-of-Field Imaging with Diffractive Optical Elements

Abstract

Today’s professional commercial compound multi-lens cameras, such as those found in smartphones, provide narrow depth of field (DoF) imaging. These devices capture images sharply at the adjusted focus point, but objects at far and close distances to the camera are extremely blurry. In this doctoral research, we aim to extend the depth of field (EDoF) in RGB imaging, achieving sharp and color distortion-free imaging across the entire imaging depth. The main goal of this thesis is to develop flexible instruments for the analysis and design of optical systems and to demonstrate the efficiency of the designed imaging optics for all-in-focus imaging, which is one of the fundamental challenges in imaging. Conventional optical engineering methods address this by designing bulky, compound, complex, and expensive multi-lens objectives. In this thesis, we consider imaging systems with two types of optical elements: lensless and hybrid.

From extensive simulation and experimental tests, we concluded that the lensless system equipped with a designed diffractive optical element (DOE) instead of a single refractive lens suffers significantly from chromatic aberration. This problem must be addressed to fully exploit the potential of DOE-based imaging architectures. To alleviate this issue, we proposed a hybrid diffractive imaging system composed of both a refractive lens and a multilevel phase mask (MPM) as a DOE.

The phase of the incident light wave is modulated based on the phase profile passing through the designed DOE. The resulting wavefront then propagates to the imaging sensor using the refractive lens to capture the data for recovery. By jointly optimizing the optics and imaging algorithms, the diffractive + refractive optical setup provides an opportunity to optimize task-specific imaging performance over traditional general-purpose cameras in various imaging applications.

In this design, we identify DOEs and the lens by the primary optical parameters: focal distance, field of view, aperture size, numerical aperture, and spectrum range. Unlike conventional optical setups using focusing refractive lenses, the following set of disjoint features are jointly optimized to produce high image quality:

1. High-fidelity phase multilevel DOEs (thickness, level of steps, aperture) produce blurred diffractive patterns oriented toward inverse image reconstruction. 2. Model-based loss function to seek the highest reconstruction quality (reference metrics, quality assessment). 3. Model of the sensor (color filter array, pitch size, demosaicing, noise models). 4. Reconstruction method (optical transfer function design, inverse imaging, learning-based deep learning). 5. Image datasets (different levels of distortions, depth range, multiwavelength images).

Moreover, a programmable phase spatial light modulator (SLM) is used as encoded DOEs along with the refractive lens to avoid building several DOEs and provide an opportunity to co-optimize the phase profile and the image reconstruction simultaneously online. The hardware-in-the-loop (HIL) methodology is investigated and implemented for the learning design of hybrid imaging, where the hardware (optics and electronics) is considered a black box embedded in the algorithmic software. The principal point is that the HIL-SLM technique will allow automatic optimal compensation for all discrepancies in the digital models, particularly for image formation, including wavefront propagation modeling through DOEs. The potential breakthrough in this approach can be found primarily in the advanced reconstruction performance. The design will also result in the optimal phase delay characteristics of DOEs, which can be exploited for manufacturing the corresponding optical elements and in the data processing algorithms enabling demosaicing, inverse imaging, denoising, super-resolution, etc.

Results indicate that the hybrid diffractive imaging system significantly improves image quality across varying depths compared to traditional multi-lens setups, demonstrating the potential for broader applications in commercial imaging technologies. The optimized hybrid optical setup achieves a focal distance of 10 mm and a 6 mm aperture, showcasing superior performance in producing colored, sharp images. Numerical and experimental analyses, including high-resolution SLM implementation, highlight the system’s advanced imaging capabilities, with superior overall imaging quality, even outperforming commercially available compound multi-lens cameras. This groundbreaking solution in miniature camera technology demonstrates the efficacy of integrating refractive and diffractive elements to achieve achromatic EDoF imaging.