/word-search-solver

Primary LanguageC++MIT LicenseMIT

This program takes an image of a word search and solves it.

There are two versions of this program: wssolve-txt and wssolve-img. They allow solving of plain text and image-based word searches, respectively. Note that the latter is more error-prone and requires a few more dependencies than the former. If this is a one-time thing, or the image is exceptionally messy, I would suggest first transcribing the word search, then using wssolve-txt.

Image Processing Dependencies

  • OpenCV 2.4: Check their website for installation instructions.
  • tesseract-ocr: Your distro's package manager should have something like tesseract-ocr or libtesseract-ocr. If not, you can always build from source.

Note that the above are only needed by wssolve-img.

Shared Dependencies:

  • cereal: For dictionary and word search data serialization.
  • libzed: Mostly just for text manipulation. See the readme for installation instructions.
  • Lastly, Make and g++ are required.

Once all dependencies are satisfied, you should be able to compile just by running make wssolve-txt and/or make wssolve-img. This should result in an executable "wssolve". To solve a word search do something like

./wssolve-img /path/to/image

or

./wssolve-txt /path/to/text/file

Notes for processing images of word searches

Accurately processing the word search currently requires there to be a border around the letters, and no other borders (the word bank is assumed to be outside of the border). This program will NOT work with oddly-shaped word searches (e.g. triangular).

For best results, follow these guidelines with images of word searches:

  • The letters in the word search should be well-aligned.
  • The letters should not have anything between them.
  • There should be no (or few) illustrations.
  • There should be nothing obscuring the letters or word bank.
  • All words in the word bank should be on the same side of the image.