Maximum-likelihood Restoration Research Articles

Pixel reassignment in image scanning microscopy with a doughnut beam: example of maximum likelihood restoration.

In image scanning microscopy, the pinhole of a confocal microscope is replaced by a detector array. The point spread function for each detector element can be interpreted as the probability density function of the signal, the peak giving the most likely origin. This thus allows a form of maximum likelihood restoration, and compensation for aberrations, with similarities to adaptive optics. As an example of an aberration, we investigate theoretically and experimentally illumination with a vortex doughnut beam. After reassignment and summation over the detector array, the point spread function is compact, and the resolution and signal level higher than in a conventional microscope.

Massively parallel spatially variant maximum-likelihood restoration of Hubble Space Telescope imagery

We present results of concurrent maximum-likelihood restoration implementations with a spatially variant point-spread function (SV-PSF) on both synthetic and real data sets from the Hubble Space Telescope. We demonstrate that SV-PSF restoration exhibits superior performance compared with restoration with a spatially invariant point-spread function. We realize concurrency on a network of Unix workstations and on a SV-PSF model from sparse point-spread function reference information by means of bilinear interpolation. We then use the interpolative point-spread function model to implement several different SV-PSF restoration methods. These restoration methods are tested on a standard synthetic Hubble Space Telescope test case, and the results are compared on a computational effort–restoration performance basis. These methods are further applied to actual Hubble Space Telescope data, including an application that corrects for motion blur, and the results are presented.

Blind image deconvolution revisited

The article discusses the major approaches, such as projection based blind deconvolution and maximum likelihood restoration, we overlooked previously (see ibid., no.5, 1996). We discuss them for completeness along with some other works found in the literature. As the area of blind image restoration is a rapidly growing field of research, new methods are constantly being developed.

Quantitative characterization of the structure of rhodopsin in disc membrane by means of Fourier transform infrared spectroscopy.

Fourier transform infrared (FTIR) spectroscopy has been used for the detailed characterization and quantification of the secondary structure of bovine rhodopsin in native disc membranes. FTIR spectra were obtained in aqueous media, both in 1H2O and in 2H2O. Analysis of spectra by means of Fourier self-deconvolution, complemented with maximum likelihood restoration and Fourier derivative, has allowed the characterization of major amide I secondary structure-sensitive component bands of structural relevance which had not been detected before. In consequence, we show a richer secondary structure for rhodopsin than previously described. Our results indicate a total regular helix content around 51%, which would include not only the main alpha 1-type helix but also 3(10)-like helix. The presence of distorted helicoid sequences might furthermore increase to a certain extent the total helix amount. It is also indicated that a significant proportion of the amino acid residues are involved in extended/beta-structures and in reverse turns, as well as in "random" segments, which had not been directly demonstrated before. 61 +/- 4% of rhodopsin is determined to be solvent-accessible, which is a substantially higher value than previously reported. Helices account for most of the inaccessible moiety.

Maximum-Likelihood Restoration Data Processing Techniques Applied to Matrix-Assisted Laser Desorption Mass Spectra

Maximum-likelihood restoration (MLR) data processing is applied to matrix-assisted laser desorption time-of-flight mass spectra of several peptide/protein samples. It is shown that this technique can enhance mass resolution for peptides (porcine insulin; MW = 5777.58 Da) sufficiently to assist in the identification of various matrix adduct ion species formed in the laser desorption process. For higher-molecular-weight proteins, mass resolution enhancement is such that peak overlap contributions from partially resolved matrix adduct ion species can be minimized. Mass accuracies are improved by approximately a factor of six for a sample of a bacterial protease (Subtilisin Carlsberg; Subtilopeptidase A) from Bacillus licheniformis (MW = 27,288.4 Da). The MLR technique also supports the contention that the mass resolution degradation observed with increasing analyte mass is at least partially due to ion signal tailing, presumably caused by secondary ion species being produced in the detection process. This factor is also postulated as being at least partially responsible for the lack of success at resolution enhancement for higher mass analytes (bovine serum albumin; MW = 66,430.3 Da) where the multiple matrix adduct ion species produced can no longer be resolved from the protonated molecular ion species, thus limiting experimental mass measurement accuracies.

Comparison of Fourier self-deconvolution and maximum likelihood restoration for curve fitting

Comparison of Fourier self-deconvolution and maximum likelihood restoration for curve fitting

Maximum-likelihood restoration of nonstationary imagery

A statistical method of extrapolating the bandwidth of nonstationary imagery is outlined. The technique combines the information that is passed by a linear optical systems, an estimate of the modulus of Fourier components in the remainder of the object spectrum, and a priori knowledge about the object to estimate the most-likely values of spectral components beyond the image bandwidth. It can be made to tolerate noise within the image bandwidth. The method is applied to images produced by objects that are finite in extent and by objects that are modulated by periodic gratings.