-
Domino Denoise: An Accurate Blind Zero-Shot Denoiser using Domino Tilings
Authors:
Jason Lequyer,
Wen-Hsin Hsu,
Reuben Philip,
Anna Christina Erpf,
Laurence Pelletier
Abstract:
Because noise can interfere with downstream analysis, image denoising has come to occupy an important place in the image processing toolbox. The most accurate state-of-the-art denoisers typically train on a representative dataset. But gathering a training set is not always feasible, so interest has grown in blind zero-shot denoisers that train only on the image they are denoising. The most accurat…
▽ More
Because noise can interfere with downstream analysis, image denoising has come to occupy an important place in the image processing toolbox. The most accurate state-of-the-art denoisers typically train on a representative dataset. But gathering a training set is not always feasible, so interest has grown in blind zero-shot denoisers that train only on the image they are denoising. The most accurate blind-zero shot methods are blind-spot networks, which mask pixels and attempt to infer them from their surroundings. Other methods exist where all neurons participate in forward inference, however they are not as accurate and are susceptible to overfitting. Here we present a hybrid approach. We first introduce a semi blind-spot network where the network can see only a small percentage of inputs during gradient update. We then resolve overfitting by introducing a validation scheme where we split pixels into two groups and fill in pixel gaps using domino tilings. Our method achieves an average PSNR increase of $0.28$ and a three fold increase in speed over the current gold standard blind zero-shot denoiser Self2Self on synthetic Gaussian noise. We demonstrate the broader applicability of Pixel Domino Tiling by inserting it into a preciously published method.
△ Less
Submitted 5 December, 2022;
originally announced December 2022.
-
Learning to segment with limited annotations: Self-supervised pretraining with regression and contrastive loss in MRI
Authors:
Lavanya Umapathy,
Zhiyang Fu,
Rohit Philip,
Diego Martin,
Maria Altbach,
Ali Bilgin
Abstract:
Obtaining manual annotations for large datasets for supervised training of deep learning (DL) models is challenging. The availability of large unlabeled datasets compared to labeled ones motivate the use of self-supervised pretraining to initialize DL models for subsequent segmentation tasks. In this work, we consider two pre-training approaches for driving a DL model to learn different representa…
▽ More
Obtaining manual annotations for large datasets for supervised training of deep learning (DL) models is challenging. The availability of large unlabeled datasets compared to labeled ones motivate the use of self-supervised pretraining to initialize DL models for subsequent segmentation tasks. In this work, we consider two pre-training approaches for driving a DL model to learn different representations using: a) regression loss that exploits spatial dependencies within an image and b) contrastive loss that exploits semantic similarity between pairs of images. The effect of pretraining techniques is evaluated in two downstream segmentation applications using Magnetic Resonance (MR) images: a) liver segmentation in abdominal T2-weighted MR images and b) prostate segmentation in T2-weighted MR images of the prostate. We observed that DL models pretrained using self-supervision can be finetuned for comparable performance with fewer labeled datasets. Additionally, we also observed that initializing the DL model using contrastive loss based pretraining performed better than the regression loss.
△ Less
Submitted 25 May, 2022;
originally announced May 2022.
-
Noise2Fast: Fast Self-Supervised Single Image Blind Denoising
Authors:
Jason Lequyer,
Reuben Philip,
Amit Sharma,
Laurence Pelletier
Abstract:
In the last several years deep learning based approaches have come to dominate many areas of computer vision, and image denoising is no exception. Neural networks can learn by example to map noisy images to clean images. However, access to noisy/clean or even noisy/noisy image pairs isn't always readily available in the desired domain. Recent approaches have allowed for the denoising of single noi…
▽ More
In the last several years deep learning based approaches have come to dominate many areas of computer vision, and image denoising is no exception. Neural networks can learn by example to map noisy images to clean images. However, access to noisy/clean or even noisy/noisy image pairs isn't always readily available in the desired domain. Recent approaches have allowed for the denoising of single noisy images without access to any training data aside from that very image. But since they require both training and inference to be carried out on each individual input image, these methods require significant computation time. As such, they are difficult to integrate into automated microscopy pipelines where denoising large datasets is essential but needs to be carried out in a timely manner. Here we present Noise2Fast, a fast single image blind denoiser. Our method is tailored for speed by training on a four-image dataset produced using a unique form of downsampling we refer to as "checkerboard downsampling". Noise2Fast is faster than all tested approaches and is more accurate than all except Self2Self, which takes well over 100 times longer to denoise an image. This allows for a combination of speed and flexibility that was not previously attainable using any other method.
△ Less
Submitted 23 August, 2021;
originally announced August 2021.
-
Methodical engineering of defects in Mn$_x$Zn$_{1-x}$ O($x$ = 0.03, and 0.05) nanostructures by electron beam for nonlinear optical applications: A new insight
Authors:
Albin Antony,
P. Poornesh,
I. V. Kityk,
K. Ozga,
J. Jedryka,
Reji Philip,
Ganesh Sanjeev,
Vikash Chandra Petwal,
Vijay Pal Verma,
Jishnu Dwivedi
Abstract:
A series of MnxZn1-xO (x=0.03, 0.05) nanostructures have been grown via the solution based chemical spray pyrolysis technique. Electron beam induced modifications on structural, linear and nonlinear optical and surface morphological properties have been studied and elaborated. GXRD (glancing angle X-ray diffraction) patterns show sharp diffraction peaks matching with the hexagonal wurtzite structu…
▽ More
A series of MnxZn1-xO (x=0.03, 0.05) nanostructures have been grown via the solution based chemical spray pyrolysis technique. Electron beam induced modifications on structural, linear and nonlinear optical and surface morphological properties have been studied and elaborated. GXRD (glancing angle X-ray diffraction) patterns show sharp diffraction peaks matching with the hexagonal wurtzite structure of ZnO thin films. The upsurge in ebeam dosage resulted in the shifting of XRD peaks (101) and (002) towards lower angle side, and increase in FWHM value. Gaussian deconvolution on PL spectra reveals the quenching of defect centers, implying the role of electron beam irradiation regulating luminescence and defect centers in the nanostructures. Irradiation induced spatial confinement and phonon localization effects have been observed in the films via micro Raman studies. The later are evident from spectral peak shifts and broadening. Detailed investigations on the effect of electron beam irradiation on third order nonlinear optical properties under continuous and pulsed mode of laser operation regimes are deliberated. Third order absorptive nonlinearity of the nanostructures evaluated using the open aperture Z-scan technique in both continuous and pulsed laser regimes shows strong nonlinear absorption coefficient \b{eta} eff of the order 10-4 cm/W confirming their suitability for passive optical limiting applications under intense radiation environments. Laser induced third harmonic generation (LITHG) experiment results supports the significant variation in nonlinearities upon electron beam irradiation, and the effect can be utilized for frequency conversion mechanisms in high power laser sources and UV light emitters.
△ Less
Submitted 8 January, 2020;
originally announced January 2020.
-
Observation of ion acceleration in nanosecond laser generated plasma on a nickel thin film under forward ablation geometry
Authors:
**to Thomas,
Hem Chandra Joshi,
Ajai kumar,
Reji Philip
Abstract:
In this article we report acceleration observed for ions in a forward ablation geometry of 50 nm thick nickel film coated on a glass plate under nanosecond laser ablation. A detailed study with varying background pressure and laser energy of time of flight (TOF) spectra of ionic and other neutral transitions from plasma is undertaken. The results indicate that TOF spectra recorded for different ne…
▽ More
In this article we report acceleration observed for ions in a forward ablation geometry of 50 nm thick nickel film coated on a glass plate under nanosecond laser ablation. A detailed study with varying background pressure and laser energy of time of flight (TOF) spectra of ionic and other neutral transitions from plasma is undertaken. The results indicate that TOF spectra recorded for different neutral transitions exhibit different dynamical behavior (the slower peak of the neutral emission becomes faster with background pressure). Similar observation is seen for the ionic species as well. These observations have similarity to some of the reported double-layer concepts. However, the calculated electric fields from the acceleration appear to be anomalously higher. Along with the acceleration observed for the ionic and neutral lines, significant asymmetry in spectral shape is observed for neutral nickel line (712.2 nm) along with spectral broadening as the background pressure increases. This large asymmetry is indicative of micro electric fields present inside the laser produced plasma plume which may result in a continuous acceleration of ions. Interestingly, the asymmetry in spectral broadening exhibits temporal and spatial dependence which indicates that electric field is present in the plasma plume even for longer duration and also over a significant distance.
△ Less
Submitted 1 May, 2020; v1 submitted 4 October, 2019;
originally announced October 2019.
-
Epitaxial Growth and Characterization of AlInN Based Core-Shell Nanowire Light Emitting Diodes Operating in the Ultraviolet Spectrum
Authors:
Ravi Teja Velpula,
Moab Rajan Philip,
Barsha Jain,
Hoang Duy Nguyen,
Renjie Wang,
Hieu Pham Trung Nguyen
Abstract:
We report on the demonstration of the first axial AlInN ultraviolet core-shell nanowire light-emitting diodes with highly stable emission in the UV wavelength range. During the epitaxial growth of AlInN layer, an AlInN shell is spontaneously formed, resulted in the reduced nonradiative recombination on nanowire surface. The AlInN nanowires exhibit high internal quantum efficiency of ~ 52% at room…
▽ More
We report on the demonstration of the first axial AlInN ultraviolet core-shell nanowire light-emitting diodes with highly stable emission in the UV wavelength range. During the epitaxial growth of AlInN layer, an AlInN shell is spontaneously formed, resulted in the reduced nonradiative recombination on nanowire surface. The AlInN nanowires exhibit high internal quantum efficiency of ~ 52% at room temperature for emission at 295nm. The peak emission wavelength can be varied from 290 nm to 355 nm by changing the growth condition. Moreover, significantly strong transverse magnetic (TM) polarized emission is recorded which is ~ 4 times stronger compared to the transverse electric (TE) polarized light at 295 nm. This study provides alternative approach for the fabrication of new type of high-performance ultraviolet light-emitters.
△ Less
Submitted 15 July, 2019;
originally announced August 2019.
-
Time-resolved optical emission spectroscopic studies of picosecond laser produced Cr plasma
Authors:
Kavya H. Rao,
N. Smijesh,
N. Klemke,
R. Philip,
I. V. Litvinyuk,
R. T. Sang
Abstract:
Time-resolved optical emission spectroscopic measurements of a plasma generated by irradiating a Cr target using 60 picosecond (ps) and 300 ps laser pulses is carried out to investigate the variation in the linewidth ($δλ$) of emission from neutrals and ions for increasing ambient pressures. Measurements ranging from 10$^{-6}$ Torr to 10$^2$ Torr show a distinctly different variation in the $δλ$ o…
▽ More
Time-resolved optical emission spectroscopic measurements of a plasma generated by irradiating a Cr target using 60 picosecond (ps) and 300 ps laser pulses is carried out to investigate the variation in the linewidth ($δλ$) of emission from neutrals and ions for increasing ambient pressures. Measurements ranging from 10$^{-6}$ Torr to 10$^2$ Torr show a distinctly different variation in the $δλ$ of neutrals (Cr I) compared to that of singly ionized Cr (Cr II), for both irradiations. $δλ$ increases monotonously with pressure for Cr II, but an oscillation is evident at intermediate pressures for Cr I. This oscillation does not depend on the laser pulse widths used. In spite of the differences in the plasma formation mechanisms, it is experimentally found that there is an optimum intermediate background pressure for which $δλ$ of neutrals drops to a minimum. Importantly, these results underline the fact that for intermediate pressures, the usual practice of calculating the plasma number density from the $δλ$ of neutrals needs to be judiciously done, to avoid reaching inaccurate conclusions.
△ Less
Submitted 21 February, 2018;
originally announced February 2018.
-
Hole Concentration Effect on the Microwave Nonlinearity of Tl2Ba2CaCu2O8-d Superconducting Thin Films
Authors:
M. M. Bischak,
J. Thomas,
R. R. Philip,
S. K. Remillard
Abstract:
The carrier concentration of Tl2Ba2CaCu2O8 films was modified by annealing in N2 gas. X-ray analysis of the structure and the oxygen content revealed a correspondence between carrier concentration and oxygen depletion. The TC and nonlinear surface impedance was measured using a dielectric resonator and the nonlinearity slope parameter r=dXS/dRS was found to converge to unity at the critical temper…
▽ More
The carrier concentration of Tl2Ba2CaCu2O8 films was modified by annealing in N2 gas. X-ray analysis of the structure and the oxygen content revealed a correspondence between carrier concentration and oxygen depletion. The TC and nonlinear surface impedance was measured using a dielectric resonator and the nonlinearity slope parameter r=dXS/dRS was found to converge to unity at the critical temperature, indicating a dominance of Josephson fluxon hysteresis on the nonlinearity. Highly inductive nonlinearity was observed in a small range of do** levels between 0.180<p<0.195 holes/Cu, which does not include the optimal do** level of 0.16 holes/Cu.
△ Less
Submitted 15 October, 2013;
originally announced October 2013.
