java - Library to segment and classify binary or grayscale images -
i interpreting scientific (stem) images component parts , adding semantics. these images born digital, noise-free , either binary (monochrome) or have small number of colours. java libraries/methods partition images whitespace-separated components , identify (classify) resulting segments. typical image is:
where want extracted segments include numerals , other characters (some rotated) , asterisks in diagram. [i use other methods extract geometrical components - e.g. bars) . library identify identical segments (e.g. 6 0 characters, 5 decimal points). have used tesseract characters many of segments may not belong unicode character set (e.g. purpose-created symbols).
update: have opened bounty. interested in libraries, not suggestions algorithms have written prototype one. if functionality part of larger system (e.g. think jbig2 has functionality) please make clear entry points are.
note: "born-digital" means image created without noise, clean lines unlike - - scanned documents.
i aware of opencv. can analyze image like:
- binarizing (if have few colors or greyscale)
- gather blobs in mat-objects
- get position of mats correct label (which should mat each letter)
- and apply algorithm mats
Comments
Post a Comment