Stroke Width Transform (Text Detection)

DIY Book Scanner Skunk Works. Share your crazy ideas and novel approaches. Home of the "3D structure of a book" thread.

Stroke Width Transform (Text Detection)

Postby aku » 26 Mar 2012, 17:11

Found the first reference at
http://www.reddit.com/r/computervision/comments/rankh/detecting_text_in_an_image_is_stroke_width/
this transform is apparently designed to help in the detection of text within images.
While this is likely not very necessary for most books, it might help with books mixing images and text, to separate them.
It might also be useful to estimate the text area on a page.

From the paper at http://yoni.wexlers.org/papers/2010TextDetection.pdf:
The SWT converts the image from containing gray values to an array containing likely stroke widths for each pixel.
This information suffices for extracting the text by measuring the width variance in each component as [...] because text tends to maintain fixed stroke width.
This puts it apart from other image elements such as foliage.


Another interesting quote, from the end of the paper, emphasis by me:
The grouping of letters can be improved by considering
the directions of the recovered strokes. This may allow the
detection of curved text lines as well.
aku
 
Posts: 27
Joined: 02 Jan 2010, 08:38
Location: Vancouver, BC

Re: Stroke Width Transform (Text Detection)

Postby daniel_reetz » 26 Mar 2012, 17:42

Very impressive - and I didn't know about /r/computervision. Thanks!
User avatar
daniel_reetz
 
Posts: 2485
Joined: 03 Jun 2009, 13:56

Re: Stroke Width Transform (Text Detection)

Postby aku » 26 Mar 2012, 19:00

daniel_reetz wrote:Very impressive - and I didn't know about /r/computervision. Thanks!


You are quite welcome.

Googling directly for the SWT showed a few followups, see below, although the original paper is still higher in the list of results. A video lecture too.

http://sites.google.com/site/roboticssaurav/strokewidthnokia
A student project implementing the SWT, claims to cover issues not mentioned in the original paper.

http://www.stanford.edu/~dmchen/documents/ICIP2011_RobustTextDetection.pdf
This one uses an euclidean distance transform as a step in computing the SWT, versus the special ray-casting of the original paper.
aku
 
Posts: 27
Joined: 02 Jan 2010, 08:38
Location: Vancouver, BC


Return to R&D and New Technologies

Who is online

Users browsing this forum: No registered users and 1 guest