Seite - 19 - in Document Image Processing

J. Imaging 2018,4, 27 Figure2.Blockdiagramof theproposedalgorithm. 2.1. TheDecisionMakingBlock Thedecisionmakingblock takesas input the image tobebinarizedandoutputsavectorwith fourparameters: thevalueof thekernel (kernel) for thebilateralfilterandthree thresholdvalues (tR, tG, tB) thatwillbe laterused in themodifiedOtsufiltering. The trainingof thebinarizationprocessproposedhere ismadewithsynthetic imageswhichwere generatedasexplained inSection2.2.Afterfiltering, thematrixofco-occurrenceprobabilitiesbetween theoriginal imageandof thebinary imagewascalculated for eachof the images in thedocument trainingset,whosegeneration isexplainedbelow. Theprobabilistic structure applied in the analysis to eachof the images in the training set is similar to the transmissionof binarydata in aBinaryAsymmetricChannel, as shown inFigure 3. The probabilities P(f/b) and P(b/f) represent an additive noise in communication channels in information theory, here it represents the inability of the algorithm to correct the back-to-front interference of the image tested in the binarization process. The probabilities P(b/b) and P(f/f) arecalculatedfromthepixel-to-pixel comparisonof thebinarizedimagegeneratedbytheproposed algorithmwith theground-truth image. Figure3.Generationof theco-occurrencematrix foreachof the images in the trainingset. 19