Empirical Determination of Experimental Error using the Twenty-Nine Duplicate Patches within the SNAP IT8.7/4r

In this paper, a new approach to visualize and calculate the unavoidable error in the duplicated patches of the SNAP IT8.7/4r characterization datasets from eighty newspaper print sites is presented. This unavoidable error is due to the spectral difference between duplicated patches in the characterization datasets. The error is linearly transformed and reported as an eigenvector/eigenvalue approximating the total error. The analysis shows that the spectral error of the duplicated patches is not similar and a systematic error bias pattern is observed. The spectral error estimation procedure includes data manipulation, bootstrapping, and principal component analysis using the R-programming language. The R-script is included as well as a script to create artificial data as an effort to stimulate further research.