Image reconstruction and enhancement by deconvolution in scatter-plate microscopy

State-of-the-Art Approaches for Image Deconvolution Problems, including Modern Deep Learning Architectures

In modern digital microscopy, deconvolution methods are widely used to eliminate a number of image defects and increase resolution. In this review, we have divided these methods into classical, deep learning-based, and optimization-based methods. The review describes the major architectures of neural networks, such as convolutional and generative adversarial networks, autoencoders, various forms of recurrent networks, and the attention mechanism used for the deconvolution problem. Special attention is paid to deep learning as the most powerful and flexible modern approach. The review describes the major architectures of neural networks used for the deconvolution problem. We describe the difficulties in their application, such as the discrepancy between the standard loss functions and the visual content and the heterogeneity of the images. Next, we examine how to deal with this by introducing new loss functions, multiscale learning, and prior knowledge of visual content. In conclusion, a review of promising directions and further development of deconvolution methods in microscopy is given.

Scatter-plate microscopy with spatially coherent illumination and temporal scatter modulation

Scatter-plate microscopy (SPM) is a lensless imaging technique for high-resolution imaging through scattering media. So far, the method was demonstrated for spatially incoherent illumination and static scattering media. In this publication, we demonstrate that these restrictions are not necessary. We realized imaging with spatially coherent and spatially incoherent illumination. We further demonstrate that SPM is still a valid imaging method for scatter-plates, which change their scattering behaviour (i.e. the phase-shift) at each position on the plate continuously but independently from other positions. Especially we realized imaging through rotating ground glass diffusers.

Imaging through opacity using a near-infrared low-spatial-coherence fiber light source

Memory-effect-based speckle correlation is one of the most practical techniques for imaging through scattering opaque media, where a light source with low spatial coherence and moderate bandwidth plays a pivotal role. Usually, a rapidly rotating diffuser is applied to make the light source spatially decoherent. Here, an all-fiber-based low-spatial-coherence light source is proposed and demonstrated for such speckle-correlated imaging. The illumination structure is greatly simplified, the lightening efficiency is enhanced, and the wavelength is extended to the near-infrared band, which is favorable for a larger memory effect range and deeper penetrating depth through opacity. Moreover, the proposed local illumination method can identify the orientation of the object, which has not been revealed by former methods. This work would facilitate the research in optical biomedical imaging and broaden the applications of multimode random fiber lasers.

Cone-shaped optical fiber tip for cost-effective digital lensless holographic microscopy

In this work, the development and application of a cost-effective and robust digital lensless holographic microscopy (DLHM) system is presented. In the simple architecture of DLHM based on a point source and a digital camera, the production of the former is introduced by means of an engineered step-index optical fiber with a cone-shaped end tip. The conventional and regularly expensive point source in DLHM is produced by means of a high-numerical-aperture microscope objective and a metallic wavelength-sized pinhole. The proposed replacement renders to DLHM additional simplicity of building, in addition to mechanical stability and robustness, and further reduces the cost of the microscope. The simplified cost-effective DLHM architecture is utilized for imaging resolution test targets and samples of human blood and pond water, revealing competitive mechanical stability and trustable phase images of the imaged specimens.

Extending the imaging range through scattering layers to the entire correlation range

A method of extending the imaging range through scattering layers around a reference point (RP) is realized. Objects within the entire correlation range of the RP can be totally recovered. By scanning the light source, objects within the memory effect (ME) range of the RP are completely recovered with high quality. By combining the shift of a camera to move the object to the center of observation view, objects far away from the RP are retrieved with an improved signal-to-noise ratio. The extended imaging range is about 3.5 times the ME range and more than 16 times the imaging range with normal static illumination. The RP can be imprecisely placed at a distance from the objects instead of precisely replacing them owing to the extended imaging range. This simple-system method forcefully breaks the limitation of the ME range and is very easy to implement in practical applications, which is meaningful for the research in scattering imaging.