Seite - 195 - in Document Image Processing

J. Imaging 2018,4, 32 Figure6.Exampleof text images fromAcTiV-Rdepictingtypical characteristicsofvideotext images. 3.2.AnnotationGuidelines Weutilized theAcTiV-GT tool [47] to annotate our collectionofdata. Figure 7 illustrates the user interfaceof this tool. In theannotationprocess,wecollect the following information for each text rectangle. • position: x,y,widthandheight. • content: text strings, textcolor,backgroundcolor,backgroundtype(transparent,opaque). • Interval: apparition intervalof the textline (Frame_S(Start),Frame_E(End)). Note thata text rectanglecan includemultiple lines if theyshare thesamefont, colorandsize, andif theyarenot far fromeachother. Caseof static text annotation: Determine (1) spatial coor. of the current text , (2) its color, (3) its bg type and color, next (4) its apparition interval List of video frames and visualization of the transcriptions' information Visualization of the layout coor. (x, y, w, h) and of the fd/bg color Figure7.AcTiV-GTopen-source tooldisplayinga labeledframe. This set of information is saved in ametafile calledglobalXMLfile (an extract is illustrated inFigure8). Thisfile canbeused for trackingandend-to-end tasks. InAcTiV2.0, twoadditional 195