Two-dimensional multifibre spectral image correction based on machine learning techniques

We propose a new method of blind deconvolution (BD) suitable for the recovery of truncated blurred images. In practical cases we often encounter truncated images, but they are difficult to treat properly with conventional BD algorithms, since nontruncated blurred images are assumed. The proposed method is based on simulated annealing (SA), to which we introduce an adaptive masking process. After a truncated segment is cut out from the blurred image, the sharp edges surrounding the segment are removed by multiplying the initial window, whose shape is determined with the initial estimate of the point spread function (PSF). A more appropriate estimate for the PSF is obtained from the masked segment based on the SA algorithm, and according to the new estimate of the PSF, the shape of the window is varied. These procedures are iterated until we finally obtain the best possible estimate of the PSF. In the last step, the entire blurred image is deconvolved with the final estimate of the PSF to obtain the recovered image. The effectiveness of the proposed method is confirmed by computer simulation.

We perform an analysis of the TRACE telescope blur from EUV images. Theblur pattern is discussed in terms of the telescope point-spread function (PSF) for the 171 A filter. The analysis performed consists of two steps. First, an initial shape for the PSF core is determined directly from TRACE EUV images. Second, the blind-deconvolution method is used for obtaining the final PSF shape. The PSF core peak is fitted by analytical functions to determine its parametric characteristics. The determined PSF includes the core central peak and peaks caused by diffraction effects inherent in TRACE EUV data. The diffraction portion of the PSF is studied theoretically in the Fraunhofer diffraction limit. The temperature dependence of the TRACE PSF shape is investigated for a selected temperature range. We also discuss general properties of the obtained PSF and its possible applications.

Blind deconvolution subject to sparse representation for fluorescence microscopy

Cone-beam CT (CBCT) realizes true three-dimensional (3D) imaging in terms of its direct volume reconstruction with isotropic resolution. However, the 3D imaging performance of a CBCT system is spatially variant (or non-uniform) over the support domain, which can be quantitatively characterized by 3D point spread function (PSF). The CBCT system PSF can be experimentally measured through the use of telfon ball and edge-spread technique. For a single circular scan orbit, its volume reconstruction fails to meet data sufficiency condition, consequently causing spatial shift variance. In the pursuit of meeting data sufficiency condition, we have proposed a circle-plus-arc CB scan scheme. The overall CBCT imaging process involves several factors, including x-ray source, cone-beam projection, and computational reconstruction; each factor can in principle be characterized by a convolution kernel or PSF. In this paper, we concentrate on the PSF characterization of circle-plus-arc algorithm. Based on the linearity of Radon transform and inverse Radon transform, we can partition the Radon domain. In particular, the circle-plus-arc scan scheme partitions the Radon domain into a donut-like region (associated with the circle scan) and funnel-like null region (provided by the arc scan). A modified FDK algorithm is responsible for donut region reconstruction, and a filtered-backprojection-styled inverse Radon transform is for the null region reconstruction. By adding them up, we obtain the complete volume reconstruction. Through the use of a bead array phantom (a 5×5×5 array), subject to cone-beam scan under different scan patterns and volume reconstruction, we calculated the local PSF blurs and the spatial variance. The result shows that, for our CBCT simulation with 28 degree cone angle, the circle scan produces a spatial variance of 9.3% and the circle-plus-arc scan reduces that to 1.8%.

A blind deconvolution problem is newly stated with the following conditions: the point-spread function is band limited, both the object and the point-spread function are nonnegative, and the solution is to be a diffraction-limited object. A blind deconvolution method was developed that can easily be applied to problems in optics because of the conditions used. The performance of the method is investigated with computer simulations.

A new blind deconvolution method with additional phase spectral constraints for the blurred image is discussed in this paper. The blurred image is a convolution of an original image and a point-spread function (PSF). The proposed method is based on the following iterative procedures, i.e., a projection onto a frequency space satisfying phase spectrum constraints, a minimization of a cost function and a projection onto an image space satisfying the support and nonnegativity constraints. This method restores the original image and the PSF stably with high accuracy. The effectiveness of the proposed method is verified by applying it to some blind deconvolution problems of images. The experimental results show that the performance of the proposed method is drastically improved in comparison with the conventional blind deconvolution method.

A scatter-glare correction technique for x-ray images acquired with an antiscatter grid was developed. In the technique, the scatter-glare image was estimated from exposure conditions and subtracted from the acquired image. The basic procedure in the estimation of the scatter-glare image is convolution filtering; however, the novel aspects of the technique are as follows: (1) To estimate the scatter-glare intensity, a formula that does not include the term of object thickness was used. With this formula, the correction can be performed, even for nonuniform phantoms; (2) To estimate the scatter-glare distribution, the experimental scatter-glare point spread function (PSF) was directly used as a convolution kernel. Although the shape of the PSF changed slightly for water thicknesses of 5-25 cm, we applied the PSF measured at a water thickness of 15 cm to the correction experiments. For the stepped water phantom (10-20 cm), scatter-glare estimation produced an average error of 10%, with respect to the lead bar data. Furthermore, the improvement of image quality and quantitative accuracy resulting from the correction was examined.

Direct exoplanet detections are limited by the speckle noise of the point spread function (PSF). This noise can be reduced by subtracting PSF images obtained simultaneously in adjacent narrow spectral bands using a multichannel camera (MCC). Experiments have shown that speckle attenuation performances are severely degraded by differential optical aberrations between channels that decorrelate the PSFs of the different spectral bands. We present a new technique which can greatly alleviate this problem: the introduction of a holographic diffuser at the focal plane of the MCC. The holographic diffuser converts the PSF image into an incoherent illumination scene that is then re-imaged with the MCC. This imaging process is equivalent to a convolution of the scene with the PSF of each channel of the MCC. The optical aberrations in the MCC then affect only the convolution kernel of each channel and not the PSF globally, resulting in more correlated images. We report laboratory measurements with a dual channel prototype (1.575 μm and 1.625 μm) to validate this approach. We achieved a speckle noise suppression factor of 12-14, which is ~4-6 times better than what has been achieved by existing MCCs.

Many astronomical studies model stars like radiating blackbodies with unit emissivity. Their conclusions should be reconsidered if the stellar spectral emissivity, ${ϵ}_{\ensuremath{\lambda}}$, were to be proven to be appreciably smaller. However, determining ${ϵ}_{\ensuremath{\lambda}}$ from raw observational data poses serious technical challenges. Here, using a machine learning technique, we implemented an inverse model for calculating the stellar spectral radiation flux in a given spectral band emissivity. Radiation flux data in some spectral bands serve as input to determine the unknown model parameters. To this purpose, we chose 411 stars (361 from the Midcourse Space Experiment (MSX) catalog and 50 from the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) catalog) as training samples of a stochastic particle swarm optimization algorithm. The mean values of the emissivity estimates thus obtained deviate significantly from the ideal blackbody value. Knowledge of the model parameters then enabled us to calculate the radiation fluxes in other spectral bands to compare with the existing observational data and thus validate our approach. Finally, based on the trained algorithm, we discuss our predictions for spectral bands where astronomical data are unavailable. Besides providing direct evidence against modeling stars as emitting blackbodies, our conclusions also call for more direct investigations of the stellar emissivity.

Shannon numbers of diffraction images

In this paper the problem of restoring an image distorted by a linear space-invariant (LST) point-spread function (PSF) which is not exactly known is solved using a maximum-a posteriori (MAP) formulation. Using the diagonalization properties of the discrete Fourier transform (DFT) for circulant matrices, the MAP estimate is computed in the DFT domain. The similarities between the MAP and the regularized constrained total least-squares (RCTLS) estimates are reported and their equivalence under certain conditions is discussed. A perturbation analysis of the MAP estimate is performed to obtain its mean-squared error (MSE) in a closed form, for small PSF errors. Numerical experiments for different PSF errors are performed to test the MAP estimator for this problem. Objective MSE-based and visual comparisons with the modified linear minimum mean-squared error (LMMSE) filter are presented that verify the superiority of the MAP estimator for higher PSF errors. >

Turbulent pockets of air in the atmosphere blur and disrupt the Point Spread Function (PSF) of a surveillance system. The PSF peak is offset in location, changed in size and in short exposures (or mild conditions) noticeably fragmented or speckled. These effects result in what is called heat shimmer or scintillation. The turbulence also causes the effective PSF convolution kernel to vary over time as well as across the image. When doing long range surveillance even mild turbulence can significantly degrade the surveillance system performance. This paper discusses an image processing method that tracks the behaviour of the PSF and then de-warps the image to reduce the disruptive effects of turbulence. Optical flow, an average image filter and a simple unsharp mask is employed to respectively track the centre of the PSF, de-speckle the image and regain the sharpness of the image. The algorithm is efficient enough to process 720p video at 10 frames per second on low power CPU-only platforms such as an Intel i5 NUC.

We propose a solution to the image deconvolution problem where the convolution kernel or point spread function (PSF) is assumed to be only partially known. Small perturbations generated from the model are exploited to produce a few principal components explaining the PSF uncertainty in a high-dimensional space. Unlike recent developments on blind deconvolution of natural images, we assume the image is sparse in the pixel basis, a natural sparsity arising in magnetic resonance force microscopy (MRFM). Our approach adopts a Bayesian Metropolis-within-Gibbs sampling framework. The performance of our Bayesian semi-blind algorithm for sparse images is superior to previously proposed semi-blind algorithms such as the alternating minimization algorithm and blind algorithms developed for natural images. We illustrate our myopic algorithm on real MRFM tobacco virus data.

Noise and blur, present in images after acquisition, negatively affect their further analysis. For image enhancement when the Point Spread Function (PSF) is unknown, blind deblurring is suitable, where both the PSF and the original image are simultaneously reconstructed. In many realistic imaging conditions, noise is modelled as a mixture of Poisson (signal-dependent) and Gaussian (signal independent) noise. In this paper we propose a blind deconvolution method for images degraded by such mixed noise. The method is based on regularized energy minimization. We evaluate its performance on synthetic images, for different blur kernels and different levels of noise, and compare with non-blind restoration. We illustrate the performance of the method on Transmission Electron Microscopy images of cilia, used in clinical practice for diagnosis of a particular type of genetic disorders.

We present the measurements of the spot size of a powerful X-ray generator at an ISAS 30m X-ray beamline facility. The generator was manufactured by Rigaku corporation. A parallel light is ideal to measure the point spread function of an X-ray telescope. However, we cannot make purely parallel light because we have to use an X-ray source with a finite size at a finite distance in a ground facility. When we evaluate a performance of X-ray telescope on the ground, an apparent size of an X-ray source often limits measurements of its point-spread function (PSF). The smaller X-ray source at the larger distance is needed to measure the PSF core of X-ray telescope. A high power X-ray source is also need to illuminate a telescope with a raster scan method that allows us precise measurements of a PSF tail. Therefore a powerful X-ray generator with a small spot size is strongly favoured in terms of the measurements of X-ray telescopes. We then introduced a powerful 18 kW rotating anode generator with a fine point focus at ISAS in 2016. The minimum and maximum voltage is 5 and 60 keV, respectively. A small spot size is technically more difficult on a lower voltage. In order to measure the spot size of the generator, we place a pinhole with a 0.1 mm diameter at 12.7 m away from the generator. The flipped image was taken at a CCD imager place at 33.5 m away from the generator. We made measurements of the Cu-K line at 8 keV with two options. One is a high-power mode (normal filament option) with a wider spot size whereas the other is a low-power mode with a finer size (fine filament option). As a result, we found that a spot size of the normal and fine filament options are 2.8 mm and 0.80 mm (FWZI), respectively. We supplied a bias voltage of 800 V for the fine filament option. The spot size of 0.8 mm is equivalent to the angular extent of as good as 6 arcsecs in the ISAS X-ray beamline facility.