(A) denoising algorithm for digital breast tomosynthesis image via patch-based low-rank penalty

Abstract

Digital breast tomosynthesis has developed to resolve the superposition problem in digital mammography. The incomplete 3-dimensional image is obtained from the reconstruction process with many projection data scanned in limited angle. During reconstruction process, the incompleteness of the 3-d image causes artifacts, but it is proven that the artifact reduction technique based on the backprojection filtration algorithm successfully reduce the artifacts. On the other hand, the noise created from the scanning process is one of the main causes that degrade the image quality. In this paper, by exploiting the characteristic of the focal plane that the locations of the lesion including microcalcification in the image are identical in the plane, we propose the denoising algorithm where the patch-based low-rank penalty is applied to the backprojected data derivatives, and compare the result of the algorithm with that of bilateral filter.