Research Highlights

From my PhD, postdoc, and side projects...

Deblurring handheld camera images

The problem of deblurring an image when the blur kernel is unknown remains challenging after decades of work. Recently there has been rapid progress on correcting irregular blur patterns caused by camera shake, but there is still much room for improvement. We propose a new blind deconvolution method using incremental sparse edge approximation to recover images blurred by camera shake. We estimate the blur kernel first from only the strongest edges in the image, then gradually refine this estimate by allowing for weaker and weaker edges. Our method competes with the benchmark deblurring performance of the state-of-the-art while being significantly faster and easier to generalize.

Slides from talk at IEEE International Conference on Image Processing, 2013

Github source and demo files


Paul Shearer, A. C. Gilbert, Alfred O. Hero III. Correcting camera shake by incremental sparse approximation. IEEE International Conference on Image Processing (2013). Outstanding Paper Award, Open Category.

"Plug-and-play" structured matrix factorization

Matrix factorization is a key signal processing and data analysis tool, used everywhere from recommender systems to bioinformatics. One limitation of present matrix factorization codes is that they only allow for a limited, prespecified set of structural constraints, most commonly the nonnegativity of entries. This code utilizes a generalized projected/proximal gradient iteration to enable the user to specify arbitrary constraints. The current code is in MATLAB but a port to Python is in progress. One application of this code is to gene microarray analysis, where it can identify people with the flu via their gene expression patterns.

Github source and gene microarray demo

IPython notebook: Diagnosing the flu from genes with scikit-learn

Classifying rare ions in the solar wind

The SWICS instruments aboard ACE and Ulysses have performed in situ measurements of individual solar wind ions for a period spanning over two decades. Solar wind composition is determined from these measurements by accumulating them into an ion count histogram, where each species appears as a distinct peak. Assigning these counts to the appropriate species is a challenging statistical problem because of the limited counts for some species and overlap between some peaks. We show that the most commonly used count assignment methods can suffer from significant bias when a highly abundant species overlaps with a much less abundant one, and bias is greatly reduced by switching to a rigorous maximum likelihood-based method. This method has been applied to reanalyze the archived ACE and Ulysses data and obtain revised abundances of rare ions in the solar wind.

Slides from Invited Talk at AGU Fall Meeting


Paul Shearer et al. The solar wind neon abundance observed with ACE/SWICS and Ulysses/SWICS. The Astrophysical Journal 789 (1), 60

Removing stray light from solar images

Extreme ultraviolet (EUV) telescopes take images of the Sun in multiple wavelengths simultaneously, and are an essential tool for astronomers to understand solar structure and dynamics. One limitation of EUV telescopes arises from slight irregularities in the surfaces of their internal mirrors, which result in scattered light and reduced contrast. In this project, point spread functions (PSFs) were determined for the extreme ultraviolet (EUV) solar imaging instruments EUVI-A and B aboard the STEREO-A and B spacecraft, and for the SWAP instrument aboard the PROBA2 mission. These PSFs enable correction of the associated EUV images for stray light contamination. The PSFs were estimated using favorable observations which provided side constraints sufficient to determine the PSFs by semi-blind deconvolution. We find that stray light contamination levels are high in faint regions, approaching 60% in coronal holes and filament cavities and often exceeding 90% in regions far outside the solar disk. Stray light correction therefore substantially reduces the inferred differential emission measure and plasma density. In some cases the correction is strongly wavelength dependent, so that the inferred plasma temperature distribution must also be revised.


Daniel B. Seaton, A. de Groof, P. Shearer, D. Berghmans, B. Nicula. SWAP observations of the long-term, large-scale evolution of the EUV solar corona. The Astrophysical Journal 777, 72 (2013).

Paul Shearer, R. A. Frazin, A. O. Hero III, A. C. Gilbert. The first stray light corrected extreme ultraviolet images of solar coronal holes. The Astrophysical Journal Letters 749.1, L8 (2012).

A. M. Vasquez, R. A. Frazin, Z. Huang, W. B. Manchester, and P. Shearer. The 3D solar minimum with differential emission measure tomography. Proceedings of the International Astronomical Union, 7, 123-133 (2011).

Copyright © 2015 Paul Shearer.