Seite - 176 - in Document Image Processing

J. Imaging 2017,3, 62 Figure3. Inkdegradationonanolddocument. (Left)original image. (Right)degradedimage. 3.2. PhantomCharacter Thephantomcharacter isa typicaldefect thatappearsondocuments thathavebeenmanually printed(usingawoodenormetal character).Aftermanyuses,aprintingcharactercanbeeroded. It is thuspossible that inkreaches thebordersof thepiece;bordersare thenprintedonthesheetofpaper. DocCreatorprovidesanalgorithmthat reproducessuchinkapparitionaroundthecharacters. Tobe asrealisticaspossible,wehavemanuallyextractedmore than30phantomdefects fromreal images. Thesedefectsare thenautomaticallyputbetweencharacters followingapatch-basedalgorithm. Thedegradationalgorithmworksas follow: (1) theuserprovidesan imageandthepercentof character todegrade; (2) charactersareextractedusingaconnectedcomponentalgorithm; (3)a list of characters is randomly set; (4) for each selected character; (4.1) a phantomdefect is randomly selectedfromthemanuallyextractedavailabledefects; (4.2) thephantomdefect is resizedtofitwith thecharactersize; (4.3) toberealistic, thephantomdefect isusedonlyasapattern; thepixelswithin thepatternare transformedusingapatchalgorithminspired from[48]whereazone fromanother partof thedocument image isselectedandcopiedwithin thepatch. SeeFigure4 foranexample. Figure4.Phantomcharacterapparition. (Left)original image. (Right)degradedimage. 3.3. PaperHoles Manyoldorrecentdocument imagescontainholes. Theseholeshavedifferentshapes, sizesand locations. DocCreator provides an algorithm that creates different kinds of holes in a document image. Thisalgorithmsimplyrandomlyappliesholesextractedfromrealdocument imagesonagiven document image. SeeFigure5 forexamples. Figure5.Cont. 176