Adaptive center determination for effective suppression of ring artifacts in tomography images

TY - JOUR

T1 - Adaptive center determination for effective suppression of ring artifacts in tomography images

AU - Jha, Diwaker

AU - Sørensen, Henning Osholm

AU - Dobberschütz, Sören

AU - Feidenhans'l, Robert Krarup

AU - Stipp, Susan Louise Svane

PY - 2014/10/6

Y1 - 2014/10/6

N2 - Ring artifacts on tomogram slices hinder image interpretation. They are caused by minor variation in the response from individual elements in a two dimensional (2D) X-ray detector. Polar space decreases the suppression complexity by transforming the rings on the tomogram slice to linear stripes. However, it requires that the center of rings lie at the origin of polar transformation. If this is not the case, all methods employing polar space become ineffective. We developed a method based on Gaussian localization of the ring center in Hough parameter space to assign the origin for the polar transformation. Thus, obtained linear stripes can be effectively suppressed by already existing methods. This effectively suppresses ring artifacts in the data from a variety of experimental setups, sample types and also handles tomograms that are previously cropped. This approach functions automatically, avoids the need for assumptions and preserves fine details, all critical for synchrotron based nanometer resolution tomography.

AB - Ring artifacts on tomogram slices hinder image interpretation. They are caused by minor variation in the response from individual elements in a two dimensional (2D) X-ray detector. Polar space decreases the suppression complexity by transforming the rings on the tomogram slice to linear stripes. However, it requires that the center of rings lie at the origin of polar transformation. If this is not the case, all methods employing polar space become ineffective. We developed a method based on Gaussian localization of the ring center in Hough parameter space to assign the origin for the polar transformation. Thus, obtained linear stripes can be effectively suppressed by already existing methods. This effectively suppresses ring artifacts in the data from a variety of experimental setups, sample types and also handles tomograms that are previously cropped. This approach functions automatically, avoids the need for assumptions and preserves fine details, all critical for synchrotron based nanometer resolution tomography.

U2 - 10.1063/1.4897441

DO - 10.1063/1.4897441

M3 - Journal article

SN - 0003-6951

VL - 105

JO - Applied Physics Letters

JF - Applied Physics Letters

M1 - 143107

ER -