PixLabeler
PixLabeler is a research tool for labeling elements in a
document image at a pixel level. PixLabeler can be used to
produce groundtruth data for training and evaluating algorithms. For
example, the foreground pixels in a bitonal image can be colored blue, cyan,
and red to represent machine print, graphics, and handwriting, respectively.
now PixLabeler Version 2.1, August, 2010
New with PixLabeler Release 2.1:
- Segmentation Mode
- Supports labeling up to 65,534 items per image.
- Auto label-advance.
- Eyedropper tool to select a visible label.
- Command line arguments for program launch.
- Toggle view the original image.
Links
- Labeled handwritten historical correspondence from
Professor Michael Brown's group at the National
University of Singapore: Binarization Shop
Eric Saund
Jing Lin
Prateek Sarkar
Michael Ivanov
Perceptual Document Analysis Area
Palo Alto Research Center
Contact: saund@parc.com
