Date of Award:
Master of Science (MS)
Nicholas S. Flann
Imagery obtained with poorly calibrated sensors is often corrupted with fixed pattern noise. Fixed pattern noise presents itself through a non-uniform distribution and therefore is hard to target in noise removal. Traditional noise removal techniques assume that the noise is uniformly distributed and subsequently produces inadequate corrections. Noise correction methods that target fixed pattern noise rely on dynamically identifying present noise and adjust correction values appropriately using nearby information or general assumptions about the image’s composition. If noise identification is not accurate, the correction values will also suffer from low accuracy. Inaccurate correction values can affect the imagery’s quality, and in some cases, produce a corrected image worse off than an uncorrected image. The proposed algorithm utilizes local and global information to find more accurate correction values on a row-by-row basis. This paper will also introduce a standard dataset and evaluation metrics for comparison against other established non-uniformity correction methods.
Imperial, Andres, "Fixed Pattern Noise Non-Uniformity Correction Through K-Means Clustering" (2021). All Graduate Theses and Dissertations. 8112.
Copyright for this work is retained by the student. If you have any questions regarding the inclusion of this work in the Digital Commons, please email us at .