Lensfree holographic microscopy takes a different approach to optical imaging: instead of forming an image with lenses, it records light transmitted through a specimen directly on a nearby digital sensor and reconstructs the image computationally. The approach can combine a compact setup with a wide field of view (FOV), making it an example of how sensor technology and image processing can expand the capabilities of on-chip microscopy.
Its central engineering challenge is to recover fine detail from measurements that are limited by sensor pixels and do not directly provide all the information needed to describe the light field. Work described in Wei Luo’s 2016 UCLA dissertation addresses these constraints through pixel super-resolution, synthetic aperture imaging, wavelength scanning and a unified reconstruction framework called the propagation phasor approach.
How lensfree holographic microscopy works
In a lensfree on-chip configuration, the sample sits close to an image sensor, such as a CCD or CMOS array. The light scattered by the specimen interferes with unscattered background light at the sensor, producing an in-line hologram. Software processes that recorded pattern to reconstruct an image of the specimen.
Because the system operates at unit magnification, the imaging FOV corresponds to the sensor’s active area. This distinguishes the architecture from conventional microscope designs, where increasing the FOV and maintaining fine spatial resolution can require different optical trade-offs. A close sample-to-sensor distance also allows partially coherent sources such as LEDs to be used in the described systems.
However, the sensor does not capture a continuous light field. Its pixels sample and integrate incoming light, so pixel pitch and the two-dimensional response of each pixel affect the recorded hologram. In holographic imaging, reconstruction must also account for diffraction between the specimen and sensor. These limitations make image processing a core part of the instrument rather than an optional post-processing step.
Pixel super-resolution addresses sensor undersampling
When a feature is smaller than the sensor’s effective sampling scale, a single exposure may not capture enough spatial information to resolve it. Pixel super-resolution addresses this by combining multiple measurements of the same specimen that contain slightly different sampling positions. The additional measurements can be produced by shifting the illumination or, in some configurations, by using sample motion.
In the approach described in the dissertation, incorporating the sensor’s two-dimensional pixel function into holographic reconstruction helped improve the numerical aperture (NA) of reconstructed images by about a factor of three compared with raw lensfree images. The work reported an NA of approximately 0.83 across an FOV of about 20 mm² using a color CMOS sensor, and an NA of approximately 0.14 across about 18 cm² using a monochrome CCD sensor. These are results from particular configurations, not general performance guarantees for every sensor or sample.
The different examples illustrate a design consideration for imaging teams: the sensor, illumination, sample spacing and reconstruction method work together. A large active sensor area can support a broad FOV, while pixel super-resolution can extract spatial detail beyond what a single raw exposure provides. The pixel response itself also matters and can be included as part of the reconstruction model.
Synthetic aperture and wavelength scanning add spatial information
Pixel super-resolution is not the only way to expand the spatial information available to reconstruction. Lensfree imaging using synthetic aperture (LISA) combines measurements made with different illumination angles. Because each angle contributes different spatial-frequency information, computationally combining the measurements can produce an effective aperture larger than that of a single illumination condition.
The dissertation reports an effective NA of approximately 1.4 across an FOV of about 20 mm² for LISA. It also describes color imaging of tissue samples, including pathology slides, using complex wave retrieval. The result depends on combining measurements and reconstruction, rather than simply inspecting one sensor exposure.
Wavelength scanning pixel super-resolution offers another measurement strategy. It uses undersampled diffraction patterns recorded at a small number of wavelengths across a narrow spectrum, with an iterative algorithm to reconstruct the specimen. Combined with synthetic aperture imaging, the described method achieved approximately 250 nm resolution, corresponding to an NA near 1.0 across an FOV greater than 20 mm². Compared with lateral-displacement-based super-resolution, wavelength scanning was reported to improve resolution uniformly in all directions and require fewer measurements.
Propagation phasors unify reconstruction steps
Conventional workflows may treat pixel super-resolution and phase retrieval as separate computational blocks. Pixel super-resolution addresses the sensor’s limited sampling, while phase retrieval estimates information about the light field needed to reconstruct the specimen’s amplitude and phase.
The propagation phasor approach combines these tasks in a unified mathematical framework. According to the dissertation, the method reduced the number of raw measurements by five to seven times compared with previous holographic reconstruction algorithms while maintaining competitive resolution across a large FOV. This data-efficiency objective is important for computational imaging systems: the reconstruction algorithm must balance image quality with the number of measurements required to obtain it.
Practical considerations for system design
For engineers and technical teams assessing lensfree imaging, the reported results point to several connected design questions:
- Sensor selection: Pixel size, active area and pixel response affect both sampling and the FOV available to the system.
- Illumination strategy: Partially coherent sources, multiple illumination angles and wavelength scanning offer different ways to encode information in measurements.
- Measurement and processing burden: More measurements can provide additional information, but reconstruction methods that use fewer exposures may improve data efficiency.
- Sample characteristics: Dense specimens can complicate holographic reconstruction because the background reference light may be distorted. The dissertation discusses multi-height imaging as one approach, although it requires additional measurements and computation.
These considerations also define limits. Lensfree transmission imaging requires light to pass through the specimen, and the dissertation identifies dense or nontransparent samples, fluorescence imaging and sensor-related angular distortions as continuing challenges. Results therefore depend on matching the imaging geometry and reconstruction method to the specimen and measurement task.
Conclusion: computation is part of the optical system
Lensfree holographic microscopy shows how sensor arrays and computational reconstruction can work together to deliver wide-field imaging without conventional imaging lenses. Pixel super-resolution tackles sensor undersampling; synthetic apertures and wavelength scanning add spatial information; and propagation phasors combine reconstruction steps to reduce measurement requirements. For system designers, the practical lesson is to evaluate illumination, sensor characteristics, sample constraints and algorithms as parts of one imaging architecture—not as independent choices.