Broadband Wide Field of View Imaging with Computational Mirrors
Vishwanath Saragadam, Niki Nezakati, Amit Roy-Chowdhury, Vivek Boominathan · 2026
Read Broadband Wide Field of View Imaging with Computational Mirrors by Vishwanath Saragadam, Niki Nezakati, Amit Roy-Chowdhury, Vivek Boominathan (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-00029.txt.
Introduction
Vishwanath Saragadam, Niki Nezakati, Amit Roy-Chowdhury, and Vivek Boominathan introduce computational mirrors for wide-field imaging across a very broad spectrum. New sensors can see from visible light into shortwave infrared at once, but glass lenses cannot focus that whole range together. Mirrors are free of that color problem, yet simple mirrors suffer field curvature and off-axis blur.
Their key observation is that different regions of the view come into focus at different distances from the mirror. Capturing a tiny focal stack of a few images lets a computational backend recover one sharp, all-in-focus picture. A new physics-inspired model, SeidelConv, describes the mirror's varying blur so it can be corrected. A first-of-its-kind fast fifty-millimeter system shows sharp results across visible, near-infrared, and shortwave bands with no refocusing.
Catalog record
| CREATOR | Vishwanath Saragadam, Niki Nezakati, Amit Roy-Chowdhury, Vivek Boominathan |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 52 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.00029
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Broadband Wide Field of View Imaging with Computational Mirrors Vishwanath Saragadam, Niki Nezakati, Amit Roy-Chowdhury, Vivek Boominathan Abstract Traditional glass-based optics are typically optimized for narrow spectral bands, such as the visible (400–700nm) or shortwave infrared (1000–1800nm). While the emergence of VIS-SWIR sensors (400–1700nm) offers transformative potential, refractive optics struggle to focus this entire range simultaneously. Mirrors represent a promising achromatic alternative; however, they are often sidelined by field curvature, and off-axis aberrations. This paper introduces Computational Mirrors , a framework that enables high-resolution, wide-field-of-view imaging across the complete VIS-SWIR spectrum using a single sensor. Our method is built on the observation that distinct regions of the field of view reach focus at varying distances from the mirror. By capturing a minimal focal stack (2–4 images), we utilize a computational backend to recover a sharp, all-in-focus image. A key contribution of this work is SeidelConv , a novel, physics-inspired, spatially-varying point spread function (PSF) model designed to accurately characterize and correct the off-axis aberrations inherent in simple concave mirrors. We demonstrate the efficacy of our approach using a first-of-its-kind 50mm F/1 optical system equipped with a VIS-SWIR sensor. Our system produces sharp images across RGB, NIR, and SWIR wavelengths without requiring refocusing, revealing material details invisible within individual spectral bands. We further validate the scalability of our approach with a 100mm F/2 system optimized for long-range imaging. 1 Introduction Figure 1 : Computational mirrors. (a) Traditional refractive (glass) optics require complex, multi-element assemblies that are inherently bulky and prone to chromatic aberration across the VIS-SWIR spectrum. In contrast, reflective optics are achromatic but suffer from severe off-axis geometric aberrations. (b) Our approach captures a sparse focal stack using simple mirror optics and employs a multi-image deconvolution backend to recover a sharp, wide-field-of-view image. (c) The proposed system enables high-performance, ultra-broadband imaging (400–1700nm) using a compact hardware footprint. (d) Long-Range Imaging: Demonstration of sharp, multi-spectral capture at significant distances across visible and NIR bands. (e) Material Analysis: An application in currency authentication, where imaging across visible and SWIR wavelengths reveals security features invisible to standard sensors. Recent advancements in sensor technology have unlocked imaging capabilities across an expansive spectral range, spanning from the visible to the short-wave infrared (SWIR) wavelengths. These broadband sensors enable high-accuracy material classification [ 5 , 24 , 4 ] , robust imaging in extreme weather conditions due to low scattering in SWIR [ 13 ] , and increased visibility in the dark [ 20 ] . However, the optics required to focus such broadband light remain a bottleneck. Traditional refractive (glass-based) optics are inherently limited by chromatic aberration, where different wavelengths focus at different depths due to the dispersive nature of glass (Fig. 1 (a)). To mitigate this, optical designers rely on complex, heavy, and often prohibitively expensive multi-element assemblies to align the broad spectral range onto a single focal plane. This complexity frequently leads to undesirable trade-offs, such as restricting the system to a narrow subset of wavelengths or bulky multi-camera architectures to cover the full spectrum. Reflective optics, including concave mirrors and complex multi-mirror objectives [ 19 ] , are inherently achromatic from the deep UV to the far-infrared. This characteristic makes them theoretically ideal for VIS-SWIR imaging. Consequently, reflective designs are standard in specialized fields such as microscopy [ 23 ] and large-scale astronomical instrumentation, most notably the James Webb Space Telescope (JWST) [ 8 ] . However, mirrors, particularly simple configurations with only one or two focusing elements, suffer from severe off-axis aberrations that constrain their versatility. These distortions typically limit the usable field of view (often to 1mm or less in microscope objectives) or necessitate high f-numbers ( f/6 or higher), which significantly reduces light throughput. One particularly debilitating effect in few-mirror systems is the Petzval (field) curvature [ 15 ] . In these systems, the image surface is non-planar; different regions across the field of view reach focus at varying distances from the mirror. This characteristic renders traditional mirror-based solutions impractical for medium focal lengths (50–100mm), where a flat sensor cannot simultaneously capture a sharp image across the entire frame. This paper presents Computational Mirrors , a practical imaging pipeline that combines hardware-controlled focal stack acquisition with a physics-motivated forward model and multi-image reconstruction. Our approach is based on two primary contributions: First, we demonstrate that while a single exposure from a simple mirror optic suffers from severe field curvature, a focal stack effectively captures sharp regions of the scene at varying axial positions. By capturing a minimal stack (2–4 images), we ensure that every point in the wide field of view is sampled at or near its focus (Fig. 1 (b)). We then employ multi-image deconvolution to fuse these measurements into a single, high-fidelity, all-in-focus image. However, these raw focal stack images are still degraded by complex off-axis aberrations. Traditional restoration methods, such as patch-wise convolution or Eigen-decomposed PSF models, are insufficient for modeling these spatial variations inherent in high-aperture mirror systems. To address this, we propose SeidelConv , a novel physics-inspired PSF model. Drawing inspiration from Spatial Transformer Networks [ 12 ] , SeidelConv accurately characterizes the spatially-varying blur across each image in the stack. SeidelConv is built on the observation that off-axis aberrations can be effectively modeled as localized geometric distortions. The input image is warped by a small set of affine transformations (typically 20 - 30) that models the distortions inherent to off-axis aberrations. These images are then blurred by learned kernels, and then weighed by per-pixel weights and summed. These parameters, including the affine matrices, blur kernels, and weight maps, are optimized using a monitor-based calibration procedure [ 21 ] . Finally, we integrate the SeidelConv forward model into a Plug-and-Play (PnP) framework [ 31 ] to recover the final sharp image, leveraging deep priors for robust regularization (Fig. 1 (c)). We validate the efficacy of Computational Mirrors through experimental prototypes using f/1 50mm and f/2 100mm concave mirror objectives. These are paired with a Sony IMX990 VIS-SWIR sensor, capable of sensing from 400nm to 1700nm. Across a diverse array of real-world scenes, ranging from macro-scale material analysis of currency to long-range outdoor environments, our system captures high-fidelity imagery across the Visible (400–700nm), NIR (700–1000nm), and SWIR (1000–1700nm) bands (Fig. 1 (d,e)). Our approach eliminates the need for wavelength-dependent refocusing, a common failure point in refractive broadband systems. These results represent the first demonstration of high-aperture ( f/1 ), ultra-broadband imaging using simple reflective optics. By overcoming the traditional limits of field curvature and off-axis aberrations, this work paves the way for a new generation of compact, lightweight, and cost-effective solutions for multi- and hyperspectral imaging. We will release the code of SeidelConv and extensive validation data for further research. 2 Prior Work Our work builds on results in aberration theory of curved surfaces such as mirrors and lenses, focal stacking and computational imaging, and broadband imaging. We explore the prior work in these areas next. Optical aberrations. Optical aberrations are broadly categorized under geometric and chromatic aberrations. Geometric aberrations occur due to light rays of a single wavelength not focusing at a single point either on axis or off-axis. These aberrations are categorized by Seidel aberrations [ 6 ] . Additionally, these optics may