Dealing with charged-coupled device noise in thin-layer videodensitometry. Optimization of several image-denoising techniques
Łukasz Komsta · Acta Chromatographica · 2009
Summary Charged-coupled device (CCD) noise can be a serious problem during videoscanning, especially when scanning dark plates with weakly fluorescent spots. The proper denoising of videoscans inside mathematical environments is a critical part of any advanced chemometric processing. The paper reports comparison and optimization of representative videoscan denoising by different techniques. Several kind of filters (averaging, circular, Gaussian, Savitzky-Golay, median, Wiener, FIR) and wavelet shrinkage (twelve mother wavelets from the Daubechies, Symmlet, and Coiflet family, five decomposition levels, and soft/hard thresholding) were optimized against noise autocorrelation or mean-squared error to the reference image. The reference image was obtained by grabbing and averaging 256 CCD frames. The median filter is the winner of the competition; other filters except Gaussian and wavelet shrinkage at high decomposition level are also sufficient and good ways of videoscan denoising. The Gaussian filter and wavelet shrinkage at low decomposition level performed worst and could not be recommended.