David Jones Brady

We show that integrated optical volume holographic devices may implement arbitrary linear transformations with order unity efficiency. At the diffraction limit, the device area needed to implement an N × N transformation is proportional to N ² λ ² , but in materials with a maximum index modulation of Δ n _max , the area scales as N ³² λ ² Δ n _max . This scaling represents a factor of a N improvement over two-port device networks, such as Mach-Zehnder arrays. A holographic system may thus implement 1000 × 1000 transformations in 30 × less area and 10, 000 × 10, 000 transformations in 100 × less area. We explore the practical implications of these scaling laws in the design of electro-optically controlled holographic vector matrix multipliers.

The standard model of photography is that lenses form images and focal planes capture those images. Over the past two decades, however, cameras have been transformed from devices that record focal images into analog-to-digital converters that transform massively parallel optical signals into serial electronic data. Under this new model, one may choose to maximize the quality and quantity of captured information, rather than focal image quality. Here we review the nature of the optical data stream and consider lens and focal plane designs that improve information capture capacity. After maximizing information capacity, the challenge of converting this information to digital data requires novel read-out and compression. We review strategies for approaching physical information limits under realistic size, power, and bandwidth constraints. Multiscale monocentric (spherical primary optics combined with arrays of secondary micro-cameras, each imaging a narrow subfield) and array lens architectures relax unfavorable geometric scaling by trading monolithic optics for co-designed system-level integration (jointly optimized optics, focal-plane, and readout subsystems). Metaoptic and mode-sorting focal plane filters implement richer projection operators that sample spectral, polarimetric, and coherence features without requiring exhaustive scanning. Finally, because readout and computation dominate energy at high throughput, we highlight architectures that perform dimensionality reduction before (or during) digitization, including integrated photonic encoders and multilayer optical projections coupled to detection and learned decoding. Together, these developments motivate a shift from cameras optimized to produce 2D images toward cameras engineered as end-to-end information channels, co-designing optics, focal-plane, and readout to deliver more task-relevant measurements per photon and per joule.

Achieving gigapixel space-bandwidth products (SBP) at video rates represents a fundamental challenge in imaging science. Here we demonstrate video-rate lensless ptychography that overcomes this barrier through the co-design of optics, sensing scheme, and computation: a coded surface encodes the dynamic object, time-sequential sensing spreads its information across frames, and a space-time neural-field framework recovers the dynamics by exploiting spatiotemporal correlations. Our approach transforms SBP scaling from sequential measurements to efficient correlation extraction. We demonstrate video-rate gigapixel imaging with centimeter-scale coverage while resolving 308-nm linewidths, achieving an SBP throughput of 20.8 gigapixels per second. Experimental validations span from monitoring mesoscale dynamics of snowflakes, bacteria, stem cells, microneedles, to characterizing time-varying dynamics in extreme-ultraviolet (EUV) experiments, demonstrating versatility across wavelengths. By transforming temporal variations from a constraint into exploitable correlations, we enable single-sensor gigapixel video that extends naturally to short-wavelength and electron regimes where radiation sensitivity has traditionally precluded high-resolution dynamic imaging.

We demonstrate a snapshot 3D and texture imaging method based on structured illumination. A calibrated projector–camera system is modeled using a differentiable projector–camera image formation model, enabling physics-guided, unsupervised learning to jointly recover 3D surface geometry and RGB reflectance from a single captured image. Experimental results show good agreement with conventional structured-light reconstruction, achieving millimeter-level depth accuracy consistent with triangulation-based resolution analysis. This method offers a promising solution for applications such as autonomous vehicle navigation and industrial metrology.

We consider sampling and detection strategies for solar-illuminated space debris. We argue that the lowest detectable debris cross-sectional area (corresponding to a factor of 3-10x in diameter) may be reduced by 10-100x by analysis of stacks of image frames collected at high rates rather than single frame data. In particular, instead of a pixel as a spatial region, the analysis is based on a "phase-space-pixel" which corresponds to an angular velocity and space region and whose intensity is computed by a weighted stacking of spatial pixels corresponding to a test debris trajectory within a wide camera field-of-view (FOV). To isolate debris signals from background, the exposure time is set to match the time it takes for debris to transit through the instantaneous field of view. Debris signatures are detected by multiscale X-ray processing of the data cube. Radiometric analysis of line integrals shows that sub-cm objects in Low Earth Orbit can be detected and assigned full orbital parameters by this approach.

Digital cameras consume $\sim 0.1$ microjoule per pixel to capture and encode video, resulting in a power usage of $\sim 20$W for a 4K sensor operating at 30 fps. Imagining gigapixel cameras operating at 100-1000 fps, the current processing model is unsustainable. To address this, physical layer compressive measurement has been proposed to reduce power consumption per pixel by 10-100×. Video Snapshot Compressive Imaging (SCI) introduces high frequency modulation in the optical sensor layer to increase effective frame rate. A commonly used sampling strategy of video SCI is Random Sampling (RS) where each mask element value is randomly set to be 0 or 1. Similarly, image inpainting (I2P) has demonstrated that images can be recovered from a fraction of the image pixels. Inspired by I2P, we propose Ultra-Sparse Sampling (USS) regime, where at each spatial location, only one sub-frame is set to 1 and all others are set to 0. We then build a Digital Micro-mirror Device (DMD) encoding system to verify the effectiveness of our USS strategy. Ideally, we can decompose the USS measurement into sub-measurements for which we can utilize I2P algorithms to recover high-speed frames. However, due to the mismatch between the DMD and CCD, the USS measurement cannot be perfectly decomposed. To this end, we propose BSTFormer, a sparse TransFormer that utilizes local Block attention, global Sparse attention, and global Temporal attention to exploit the sparsity of the USS measurement. Extensive results on both simulated and real-world data show that our method significantly outperforms all previous state-of-the-art algorithms. Additionally, an essential advantage of the USS strategy is its higher dynamic range than that of the RS strategy. Finally, from the application perspective, the USS strategy is a good choice to implement a complete video SCI system on chip due to its fixed exposure time. Code is available at https://github.com/mcao92/BSTFormer.

Synthetic aperture imaging has enabled breakthrough observations from radar to astronomy. However, optical implementation remains challenging due to stringent wavefield synchronization requirements among multiple receivers. Here we present the multiscale aperture synthesis imager (MASI), which utilizes parallelism to break complex optical challenges into tractable sub-problems. MASI employs a distributed array of coded sensors that operate independently yet coherently to surpass the diffraction limit of single receiver. It combines the propagated wavefields from individual sensors through a computational phase synchronization scheme, eliminating the need for overlapping measurement regions to establish phase coherence. Light diffraction in MASI naturally expands the imaging field, generating phase-contrast visualizations that are substantially larger than sensor dimensions. Without using lenses, MASI resolves sub-micron features at ultralong working distances and reconstructs 3D shapes over centimeter-scale fields. MASI transforms the intractable optical synchronization problem into a computational one, enabling practical deployment of scalable synthetic aperture systems at optical wavelengths.

Synthetic aperture imaging has enabled breakthrough observations from radar to astronomy. However, optical implementation remains challenging due to stringent wavefield synchronization requirements among multiple receivers. Here we present the multiscale aperture synthesis imager (MASI), which utilizes parallelism to break complex optical challenges into tractable sub-problems. MASI employs a distributed array of coded sensors that operate independently yet coherently to surpass the diffraction limit of single receiver. It combines the propagated wavefields from individual sensors through a computational phase synchronization scheme, eliminating the need for overlapping measurement regions to establish phase coherence. Light diffraction in MASI naturally expands the imaging field, generating phase-contrast visualizations that are substantially larger than sensor dimensions. Without using lenses, MASI resolves sub-micron features at ultralong working distances and reconstructs 3D shapes over centimeter-scale fields. MASI transforms the intractable optical synchronization problem into a computational one, enabling practical deployment of scalable synthetic aperture systems at optical wavelengths.

Digital cameras consume ~0.1 microjoule per pixel to capture and encode video, resulting in a power usage of ~20W for a 4K sensor operating at 30 fps. Imagining gigapixel cameras operating at 100-1000 fps, the current processing model is unsustainable. To address this, physical layer compressive measurement has been proposed to reduce power consumption per pixel by 10-100X. Video Snapshot Compressive Imaging (SCI) introduces high frequency modulation in the optical sensor layer to increase effective frame rate. A commonly used sampling strategy of video SCI is Random Sampling (RS) where each mask element value is randomly set to be 0 or 1. Similarly, image inpainting (I2P) has demonstrated that images can be recovered from a fraction of the image pixels. Inspired by I2P, we propose Ultra-Sparse Sampling (USS) regime, where at each spatial location, only one sub-frame is set to 1 and all others are set to 0. We then build a Digital Micro-mirror Device (DMD) encoding system to verify the effectiveness of our USS strategy. Ideally, we can decompose the USS measurement into sub-measurements for which we can utilize I2P algorithms to recover high-speed frames. However, due to the mismatch between the DMD and CCD, the USS measurement cannot be perfectly decomposed. To this end, we propose BSTFormer, a sparse TransFormer that utilizes local Block attention, global Sparse attention, and global Temporal attention to exploit the sparsity of the USS measurement. Extensive results on both simulated and real-world data show that our method significantly outperforms all previous state-of-the-art algorithms. Additionally, an essential advantage of the USS strategy is its higher dynamic range than that of the RS strategy. Finally, from the application perspective, the USS strategy is a good choice to implement a complete video SCI system on chip due to its fixed exposure time.