[Influence of image process on fractal morphology characterization of NAPLs vertical fingering flow]

Dyes are frequently used to visualize fingering flow pathways, where the image process has an important role in the result analysis. The theory of fractal geometry is applied to give quantitative description of the stain patterns via image analysis, which is helpful for finger characterization and prediction. This description typically involves two parameters, a mass fractal dimension (D(m)) relative to the area, and a surface fractal dimension (D(s)) relative to the perimeter. This work detailed analyzes the influence of various choices during the thresholding step that transformed the origin color images to binary ones which are needed in the fractal analysis. One hundred and thirty images were obtained from laboratory two-dimension sand box infiltration experiments of four dyed non-aqueous phase liquids. Detailed comparisons of D(m) and D(s) were made respectively, considering a set of threshold algorithms and the filling of lakes. Results indicate that adjustments of the saturation threshold influence are less on both D(m) and D(s) in the laboratory experiments. The brightness threshold adjustments decrease the D(m) by 0.02 and increase the D(s) by 0.05. Filling lakes influence the D(m) less while the D(s) decrease by 0.10. Therefore the D(m) was recommended for further analysis to avoid subjective choices' influence in the image process.