🔬 optics

Single-shot incoherent imaging with extended and engineered field of view using coded phase apertures

This paper proposes and validates a novel single-shot incoherent imaging technique that utilizes a coded phase mask to multiplex multiple isolated object areas into a camera's limited sensor region, enabling an extended and engineered field of view without compromising magnification.

Original authors: Sai Deepika Sure, Jawahar Prabhakar Desai, Joseph Rosen

Published 2026-02-06

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Sai Deepika Sure, Jawahar Prabhakar Desai, Joseph Rosen

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you have a high-powered camera lens that can see tiny details very clearly (high magnification). However, because the camera's "eye" (the sensor) is small, it can only see a tiny patch of the world at once. If you try to look at a wide scene, the edges get cut off, just like trying to see a whole mountain through a drinking straw.

Usually, to see more, you'd need a bigger camera sensor (expensive) or a special wide-angle lens (which makes the image look warped and blurry at the edges). This paper introduces a clever trick to see a much wider area without changing the lens or buying a bigger camera.

Here is how the researchers did it, explained through simple analogies:

1. The Problem: The "Tunnel Vision" Camera

Think of your camera sensor as a small window. If you are looking at three friends standing far apart in a field, your window might only be big enough to see the friend in the middle. The other two friends are standing in the "blind spots" outside the window. Normally, you can't see them.

2. The Solution: The "Magic Code" Mask

The researchers placed a special, invisible "code" (called a Coded Phase Mask) in front of the lens. Think of this mask like a prism made of invisible glass that doesn't just bend light, but scatters it in a very specific, pre-planned pattern.

The Multiplexing Trick: The mask is actually a combination of several different "scattering patterns" glued together.
The Effect: When light from a friend standing outside your window hits this mask, the mask doesn't let the image disappear. Instead, it takes that "lost" friend and creates multiple tiny, scattered copies of them (like a sparse pattern of dots) and projects them right into the center of your small window.

3. The Result: A Puzzle of Dots

When the camera takes a picture, it doesn't see a clear, wide image. Instead, it sees a strange, scattered mess of dots.

The friend in the middle looks like a normal dot pattern.
The friend who was "outside" the view now appears as a different pattern of dots, shifted to a specific spot on the sensor.
The friend on the other side appears as a third unique pattern of dots.

It's like looking at a puzzle where every piece of the wide scene has been broken into tiny, scattered shards and rearranged onto a small table.

4. The Reconstruction: The Computer "Decoder"

Since the camera only recorded these scattered dots, the image looks useless to the human eye. But the computer knows the "secret code" (the exact pattern of the mask).

The researchers use a mathematical process called deconvolution. Think of this as a super-smart puzzle solver.

The computer knows exactly how the mask scattered the light.
It takes the scattered dots on the small sensor.
It mathematically "unscrambles" them, moving the scattered dots back to their original positions.
The Magic: Suddenly, the computer reconstructs the full, wide image with all three friends visible, in their correct places, with the same high clarity as the original zoomed-in lens.

5. The Experiments

The team tested this in two ways:

Two Friends: They set up a system to see two objects that were too far apart for the camera to see at once. By using the mask, they successfully "brought" the second object into view and reconstructed the full scene.
Three Friends: They did the same thing with three objects, proving the system can handle even more "lost" areas.

They also tested how many "dots" (scattering points) worked best. They found that having a specific number of dots (like 7 or 8) made the final picture the clearest and the least noisy.

The Bottom Line

This paper shows a way to engineer the field of view. Instead of physically moving the camera or using a bigger sensor, they use a coded mask to "shrink" a wide scene into a small sensor as a pattern of dots, and then use math to "expand" it back out.

Key Takeaways from the paper:

Single Shot: You only need to take one picture. You don't need to move the camera or take multiple photos and stitch them together.
No Magnification Loss: You get the wide view without losing the ability to see small details (high magnification).
Noise Trade-off: The more areas you try to capture at once, the more "noise" (static) appears in the final image, so there is a limit to how many "friends" you can see at once.

In short, they turned a small camera window into a wide-angle lens by using a code to scatter the light and a computer to put the pieces back together.

Technical Summary: Single-shot Incoherent Imaging with Extended and Engineered Field of View Using Coded Phase Apertures

Problem Statement
Optical imaging systems, particularly those requiring high magnification, are frequently constrained by a limited field of view (FOV) due to the finite size of camera sensors. Traditional solutions to extend the FOV, such as fisheye lenses, panoramic stitching, or multiple sensor arrays, introduce significant drawbacks including severe peripheral distortions, increased system size and complexity, high costs, and the need for multiple exposures or mechanical scanning. While computational imaging techniques like Fourier ptychography offer improvements, they often require iterative reconstruction and multiple frames, limiting real-time applicability. This study addresses the challenge of expanding the FOV of a single optical system without compromising magnification or requiring multiple exposures, specifically targeting compact platforms where sensor size is a primary constraint.

Methodology
The proposed technique utilizes a Coded Phase Mask (CPM) integrated into an incoherent imaging system to engineer the field of view. The core concept involves recording an Object Response Pattern (ORP) on a camera sensor by introducing a CPM in front of the first lens (L1). The CPM is a multiplexed mask comprising $N$ distinct scattering phases, where $N-1$ represents the number of isolated object areas to be brought into the field of view beyond the system's inherent limits.

Optical Setup: The system consists of an object plane, a diffractive lens (L1) combined with the CPM, a second lens (L2), and a camera. The imaging condition is satisfied between the object and image sensor planes.
CPM Design: Each of the $N$ scattering phases is synthesized using a modified Gerchberg–Saxton algorithm (GSA). Each phase is designed to generate a unique, sparse dot pattern (Point Spread Function or PSF) in the Fourier plane (camera sensor) for a point source located in its corresponding object area. Crucially, the phases are engineered such that even if an object point lies outside the system's natural FOV, its response is shifted into the camera's sensing area.
Multiplexing: To display multiple phases simultaneously on a Spatial Light Modulator (SLM), the authors employ a spatial multiplexing technique. A random binary matrix is divided into $N$ sections, and each section selects the corresponding CPM phase. These are summed to create a single multiplexed phase mask, which is combined with the phase of the diffractive lens.
Reconstruction: The camera records a superposition of the object responses. To reconstruct the original object plane, the recorded ORP is zero-padded to the size of the magnified object plane. The system's PSFs are also zero-padded and shifted to their respective positions in a large matrix. The final image is recovered via a single-step Wiener deconvolution of the zero-padded ORP with the shifted sum of the PSFs.

Key Contributions

FOV Engineering without Magnification Loss: The study demonstrates a method to capture $N$ distinct object regions simultaneously on a single sensor without altering the system's magnification.
Single-Shot Operation: Unlike scanning-based or multi-frame computational methods, this approach captures the extended FOV in a single exposure, making it suitable for dynamic scenes.
Multiplexed Phase Mask Synthesis: The paper details a specific method for synthesizing and multiplexing $N$ scattering phases on a single SLM to create unique, sparse PSFs that allow for the separation and reconstruction of multiple object areas.
Experimental Validation: The method is validated through both simulations and physical experiments for $N=2$ and $N=3$ object areas, utilizing USAF 1951 targets.

Results

Simulations: MATLAB simulations were conducted for $N=3$ objects with a camera sensor size of $500 \times 500$ pixels and an object plane of $1080 \times 1080$ pixels. The study evaluated the Signal-to-Noise Ratio (SNR) for varying numbers of sparse dots ( $K$ ) in the PSF. The results indicated that a PSF with $K=11$ dots yielded the highest SNR.
Experiments ( $N=2$ ): Using two separate illumination channels, the system successfully imaged two objects, one of which was outside the natural FOV. Metrics including SNR, Structural Similarity Index Measure (SSIM), and Mean Squared Error (MSE) were analyzed. The optimal performance was achieved with $K=8$ dots, providing the best SSIM and lowest MSE.
Experiments ( $N=3$ ): In a single-channel setup, three objects were imaged. The FOV was computationally restricted to simulate a smaller sensor, showing only the central object in direct imaging. With the CPM, all three objects were recovered. The optimal sparse pattern was found to be $K=7$ dots, which yielded the maximum SNR, minimum MSE, and highest SSIM.
Noise Analysis: The authors note that the reconstruction process introduces background noise primarily from cross-deconvolution between different object responses ( $ORP_m$ and $PSF_n$ where $m \neq n$ ) and side lobes from intra-deconvolution.

Significance and Claims
The paper claims that this CPM-based multiplexing approach offers a distinct balance of compactness, efficiency, and single-shot operation compared to existing FOV extension techniques. It asserts that the method enables the reconstruction of multiple object regions from a single shot without mechanical scanning, preserving the native magnification and resolution of the imaging system. The reconstruction is described as efficient and non-iterative, relying on a single deconvolution step.

The authors acknowledge limitations, specifically that the number of multiplexed regions is constrained by the system aperture, leading to an SNR trade-off as the multiplexing order increases. They also identify cross-deconvolution noise as a primary source of degradation, noting that future work will investigate time-multiplexing methods to eliminate this noise, albeit at the cost of temporal resolution. The study positions this work as an extension of the I-COACH (Interferenceless Coded Aperture Correlation Holography) framework, demonstrating its capability to expand imaging limits beyond the sensor's physical dimensions.