Tool for extracting glyph images from a page xml file for an ocr image.
#Usage ImageExtractor -i inputImageFile -x pageXml -o outputFolder
-i: Path to the ocr image from which the glyphs will be extracted.
-x: Path to the aletheia page xml file.
-o: Path to the folder to put all glyph images in.
#Description Aletheia is a tool for creating ground truth for an ocr image. It helps users to correct ocr scans for a given image so the user can define boxes for each scanned character and its correct Unicode character. These boxes and characters are saved in a so called page.xml.
Franken+ is a tool which parsed this page.xml to create a new font for training tesseract. This font consists of the glyph defined by the bounding boxes in the page.xml.
#Problem Franken+ uses an external tool for extracting the actual glyph images from the ocr image. This tool only uses the bounding boxes to extract the glyph although Aletheia allows to define bounding polygons for glyph. For most fonts and characters the bounding boxes is sufficient, but for some characters it creates glyphs which contains parts from other glyphs. For example, the long s is a character which bounding box contains a part of the next character:
This will be extracted as:
Aletheia addresses this problem by providing the user with a tool which lets the user create a bounding polygon (the green line around the long s in the picture is such a bounding polygon whereas the q has a normal bounding box).
#Solution This image extractor uses the polygon to extract the glyph and therefor creates a clean glyph:
