Back to publications
A study of why we need to reassess full reference image quality assessment with medical images
Paper Details
Published: 2024/09/24
DOI: 10.48550/arXiv.2405.19097
ARXIV ID: 2405.19097v2
The paper highlights that the commonly used metrics for image quality assessment, SSIM and PSNR, do a poor job at selecting the best medical images in 8 different clinical scenarios. These metrics may be of use for natural images, but optimizing their performance for medical images is undesirable, in general.
Authors
A. Breger
M.S. Landman
I. Selby
N. Amberg
E. Brunner
J. Grohl
S. Hatamikia
L. Ning
C. Karner