arXiv subjects

Parker Liu

Publications and source records attributed to Parker Liu.

3 recordsLinked to original sources

arXiv2026

Lensless wide-field 3D fiber endoscopy through scattering media using synthetic wavelength holography

Minimally invasive imaging with fiber optic endoscopes is crucial for in vivo visualization of tissue morphology, as it supports applications such as early detection of tumors. However, imaging performance of conventional fiber endoscopes is limited when scattering layers are present between target and the distal end of endoscope. This limitation is particularly relevant in biomedicine, where targets such as early stage lesions or blood clots can be obscured by scattering tissue. To address this challenge, we present a lensless endoscopic imaging approach based on synthetic wavelength holography (SWH). SWH is a computational imaging technique in which two optical fields acquired at closely spaced wavelengths are combined to synthesize a field at a longer synthetic wavelength. As the field at longer wavelengths is less sensitive to path length perturbations, this approach can enable endoscopic recovery of holographic information despite scattering in the intervening tissue. In addition, because the synthetic field is assembled from scattered optical fields with larger optical \'etendue, our approach can extend the field of view (FoV) beyond the numerical aperture of the fiber. In this paper, we present the first demonstration of an SWH-based lensless endoscope using a multicore multimode fiber. We experimentally recover three-dimensional images of objects hidden behind scattering layers and through real biological tissue, with a spatial resolution of $\approx 500~\mu m$. We further demonstrate recovery of object information over an extended FoV of 46{\deg} without any distal optics. These results suggest a practical path toward extending fiber endoscopy for wide-field, three-dimensional imaging through scattering media.

physics.optics↗

arXiv2025

IR3D-Bench: Evaluating Vision-Language Model Scene Understanding as Agentic Inverse Rendering

Vision-language models (VLMs) excel at descriptive tasks, but whether they truly understand scenes from visual observations remains uncertain. We introduce IR3D-Bench, a benchmark challenging VLMs to demonstrate understanding through active creation rather than passive recognition. Grounded in the analysis-by-synthesis paradigm, IR3D-Bench tasks Vision-Language Agents (VLAs) with actively using programming and rendering tools to recreate the underlying 3D structure of an input image, achieving agentic inverse rendering through tool use. This "understanding-by-creating" approach probes the tool-using generative capacity of VLAs, moving beyond the descriptive or conversational capacity measured by traditional scene understanding benchmarks. We provide a comprehensive suite of metrics to evaluate geometric accuracy, spatial relations, appearance attributes, and overall plausibility. Initial experiments on agentic inverse rendering powered by various state-of-the-art VLMs highlight current limitations, particularly in visual precision rather than basic tool usage. IR3D-Bench, including data and evaluation protocols, is released to facilitate systematic study and development of tool-using VLAs towards genuine scene understanding by creating.

cs.CV↗

arXiv2025

Fiber Endoscopy Using Synthetic Wavelengths for 3D tissue imaging

Fiber-based endoscopes utilizing multi-core fiber (MCF) bundles offer the capability to image deep within the human body, making them well-suited for imaging applications in minimally invasive surgery or diagnosis. However, the optical fields relayed through each fiber core can be significantly affected by phase scrambling from height irregularities at the fiber ends or potential multi-mode cores. Moreover, obtaining high-quality endoscopic images commonly requires the fiber tip to be placed close to the target or relies on the addition of a lens. Additionally, imaging through scattering layers after the fiber tip is commonly not possible. In this work, we address these challenges by integrating Synthetic Wavelength Imaging (SWI) with fiber endoscopy. This novel approach enables the endoscopic acquisition of holographic information from objects obscured by scattering layers. The resulting endoscopic system eliminates the need for lenses and is inherently robust against phase scrambling caused by scattering and fiber bending. Using this technique, we successfully demonstrate the endoscopic imaging of features approximately 750micrometers in size on an object positioned behind a scattering layer. This advancement represents significant potential for enabling spatially resolved three-dimensional imaging of objects concealed beneath tissue using fiber endoscopes, expanding the capabilities of these systems for medical applications.

physics.optics↗