Source-linked AI summary
Lensless Imaging by Compressive Sensing
Gang Huang, Hong Jiang, Kim Matthews, Paul Wilford
TL;DR
The paper addresses lens-based compressive cameras that add physical image formation, size, cost, and complexity. It proposes a lensless architecture using controllable aperture transmittances and a single sensor for compressive measurements. A low-cost prototype demonstrates feasibility, while reconstruction is shown with 12.5% and 25% of total measurements.
Problem
Existing compressive cameras use lenses that constrain geometric and radiometric mapping and add size, cost, and complexity.
Method
The architecture controls a two-dimensional aperture array according to sensing-matrix rows and uses a single sensor to integrate the resulting rays.
Results
A low-cost prototype demonstrates feasibility, with reconstructed images reported using 12.5% and 25% of total measurements.
Takeaways & Limitations
The architecture supports simple, reliable lensless imaging across visible, infrared, and millimeter-wave spectra, with potential surveillance applications.
Takeaways & Limitations
The experiments use one sensor, while multi-view results with two sensors are deferred to future work.
Abstract
from arXiv · showhide
In this paper, we propose a lensless compressive imaging architecture. The architecture consists of two components, an aperture assembly and a sensor. No lens is used. The aperture assembly consists of a two dimensional array of aperture elements. The transmittance of each aperture element is independently controllable. The sensor is a single detection element. A compressive sensing matrix is implemented by adjusting the transmittance of the individual aperture elements according to the values of the sensing matrix. The proposed architecture is simple and reliable because no lens is used. The architecture can be used for capturing images of visible and other spectra such as infrared, or millimeter waves, in surveillance applications for detecting anomalies or extracting features such as speed of moving objects. Multiple sensors may be used with a single aperture assembly to capture multi-view images simultaneously. A prototype was built by using a LCD panel and a photoelectric sensor for capturing images of visible spectrum.
1. INTRODUCTION
The paper proposes lensless compressive imaging using a controllable aperture assembly and sensor, avoiding physical image formation before capture. The architecture targets simpler imaging across spectra and supports potential surveillance and multi-view applications.
- 1. INTRODUCTION: Lenses constrain scene-to-image mapping and add size, cost, and complexity to compressive cameras.
- 1. INTRODUCTION: Lensless compressive imaging uses an aperture assembly with independently controllable transmittance and a single sensor to implement sensing-matrix measurements.The prototype uses an LCD panel and a three-color photoelectric detector.
- 1. INTRODUCTION: Unlike related cameras, the proposed architecture forms no physical image before digital capture.
- 1. INTRODUCTION: The design is intended to avoid defocus blur, support multiple sensors for simultaneous multi-view imaging, and operate across visible, infrared, and millimeter-wave spectra.
- 1. INTRODUCTION: The paper identifies anomaly detection and moving-object speed extraction as potential surveillance uses.
2. DESCRIPTION OF ARCHITECTURE
The architecture maps controllable aperture patterns to compressive measurements with a single detection element. Each sensing-matrix row sets aperture transmittances, and the sensor integrates the resulting transmitted rays.
- 2. DESCRIPTION OF ARCHITECTURE: The system combines a two-dimensional controllable aperture array with a single detection element.
- 2.1 Compressive measurements: Opening aperture elements one by one produces conventional pixel measurements, while compressive sensing uses fewer patterned measurements.
- 2.1 Compressive measurements: Each sensing-matrix row defines an aperture pattern and one sensor measurement, usually using fewer rows than aperture elements.
- 2.1 Compressive measurements: A sensor measurement is the projection of the image onto a sensing-matrix row after aperture transmittances modulate and integrate the rays.
- 2. DESCRIPTION OF ARCHITECTURE: The aperture assembly can use liquid-crystal sheets for visible imaging or micromirror arrays for visible and infrared imaging.
3. RELATED WORK
Related cameras form and pixelize a physical analog image, whereas this architecture captures projections without explicitly forming a planar image. This removes defocus blur as a physical image-formation artifact under ideal components.
- 3. RELATED WORK: The proposed architecture differs from single-pixel and lensless cameras by avoiding physical image formation before pixelization or compressive capture.
- 3. RELATED WORK: Earlier cameras can suffer quality, sharpness, and resolution effects from both analog image formation and pixelization.
- 3. RELATED WORK: Because no planar image is explicitly formed, the proposed system's virtual image is free of defocus blur when the aperture assembly and sensor are theoretically perfect.
4. MULTI-VIEW IMAGING
Multiple sensors can share one aperture assembly to capture multi-view images simultaneously or combine measurements for reconstruction under suitable scene and sensor configurations.
- For a fixed transmittance pattern, each sensor records a measurement simultaneously, producing separate measurement vectors for independent image reconstruction.
- Correlations between images, especially for nearby sensors and distant scenes, can be exploited to enhance reconstructed-image quality.
- Multiple sensors with one aperture assembly capture different views of the same scene, enabling simultaneous multi-view imaging.
- When the scene is planar or sufficiently distant, sensor measurements can be concatenated to increase measurements for reconstructing the same image.
- With suitable sensor positioning in planar or distant scenes, multiple measurements can represent a higher-resolution pixelized image.
5. PROTOTYPE
The prototype uses a transparent monochrome LCD as a programmable aperture assembly and a photovoltaic sensor to acquire compressive measurements for lensless image reconstruction. Laboratory examples demonstrate reconstructions from fractional measurement sets, while multi-view experiments remain future work and color channels are uncalibrated.
- 5. PROTOTYPE: The prototype uses a transparent monochrome LCD for programmable aperture patterns and a photovoltaic sensor for light measurements.
- 5.1 Image acquisition: A computer generates aperture patterns from measurement-matrix rows, synchronizes display and capture, and records the sensor measurements.
- 5.1 Image acquisition: The LCD provides 65,534 binary aperture elements, and the prototype can acquire 65,534 distinct measurements corresponding to the image pixels.
- 5.1 Image acquisition: The reported experiments use one sensor, so multi-view imaging with two sensors is deferred to future work.
- 5.2 Image Reconstruction: Reconstructed laboratory images use total-variation L1 minimization, with 12.5% measurements for a soccer ball and 25% for books and a cat.
- 5.2 Image Reconstruction: Color images use the three directly measured color components without calibration to balance them.
6. CONCLUSION
The proposed lensless compressive imaging architecture enables simple, reliable devices without lens-induced defocus blur. A low-cost prototype demonstrates feasibility, while reconstructed images use 12.5% or 25% of total measurements.
- The architecture supports simple, reliable imaging devices with reduced size, cost, and complexity.
- Devices based on the architecture may support surveillance tasks such as anomaly detection and moving-object speed extraction.
- A low-cost prototype built from commercially available components demonstrates that the proposed architecture is feasible and practical.
- The soccer reconstruction uses 12.5% of total measurements.
- The books and sleeping-cat reconstructions each use 25% of total measurements.