Possible Duplicate:
How to extract traffic signs from a photograph?

I am working on a project which involves detecting north american speed signs (black/white) using a 720p HD camera. My problem is that I cannot seem to figure out how to classify the rectangular profile of the speed signs in order to pass it on to the text detection stage. My current segmentation approach detects differing levels of white in an image and thresholds the image based on those values (determined from color of road signs). The problem is that I need to segment it further as threshold-ed shape doesn't contain the complete road sign, but bits and pieces of it. I wanted to know what would be the best way to classify the road sign?

These are some pre-segmented images and the corresponding segmented images: enter image description here http://imgur.com/a/Hwpwq

As you can see the road sign has holes in it but its distinct rectangular profile is in-tact in most cases. I am currently looking into using kurtosis shape profile to determine if it is rectangle however the approach seems complicated and computationally expensive, I wanted to know if anyone knows a faster simpler method to determine the rectangular profile of the road sign from the segmented image?

I did this in Matlab using the image processing toolbox.

EDIT: Use Local OTSU thresholding it will allow for segmentation and further morph operations allow to speed sign detection.

Whoever marked this similar to the link below as a duplicate, its not... there is fundamental difference between chromatic and achromatic colors (This question deals with achromatic) whereas the link suggested below deals with chromatic.

Road Sign Images

More road sign images

  • $\begingroup$ Take a look at [this answer]. It's my answer to a similar question, the road signs were not black and white speed signs, but the approaches presented in the links are applicable to your domain as well. $\endgroup$
    – penelope
    Jan 2, 2013 at 13:06
  • 3
    $\begingroup$ @penelope I do not think the link is there... $\endgroup$
    – Spacey
    Jan 2, 2013 at 14:24
  • $\begingroup$ Sorry, forgot to paste in the link. It's here $\endgroup$
    – penelope
    Jan 2, 2013 at 20:15
  • $\begingroup$ Have you tried using an infrared camera ? $\endgroup$
    – rold2007
    Jan 3, 2013 at 20:24

1 Answer 1


Your question is broadly answered here: How to extract traffic signs from a photograph?

Your specific case is a little more complicated because the signs are not particularly distinguishing, however the same suggested techniques of cross correlating the images with clean images of original signs (you will probably want a library of images containing all signs for best results) are still applicable.

One suggestion above was to use an infrared camera which will be very effective if the road signs are micro prismatic or glass bead designs.


Not the answer you're looking for? Browse other questions tagged or ask your own question.