Quantification of Banding, Streaking and Grain in Flat Field Images.

Abstract

In this paper, we discuss techniques for the quantification of random and periodic noise present in flat field images. The types of noise studied are: (1) One-dimensional periodic noise (banding); (2) One-dimensional random noise (streaking) and, (3) Two-dimensional random noise (grain). A spectral separation technique was developed which allows the estimation of the individual noise power spectrum (NPS) of each artifact, based on the original flat field data. We proceed by assuming that the components are additive. First, the one-dimensional periodic components are identified, quantified, and removed from the flat field scan. Then the residual one-dimensional and two-dimensional random noise components are estimated via analysis of the twodimensional NPS. This technique is useful in the characterization of digital cameras, printers, and scanners.