loading
This Thread  New Post   Forum Index 

 
Hough Transform - Angles
Anonymous 		10 year
I've read other posts about Hough angle limits but I am still confused.  The documentation for the module does not describe the angle limits.  If the line is parameterized in angle of the normal and radius from the origin as in the first tutorial, I would expect angle limits to be between 90 degrees and -90 degrees, never more than 180.  But perhaps some modified representation is used.  It also seems no limits on radius are offered.  I am not sure if the angle limits would help much if they are in the standard representation, what I was trying to do is limit angles to be + or - 30 degrees relative to the y axis.  I'd like to get more explanation about the angles.

It seems that the module may not do well if lines are thin and even slightly curved, so using it in conjunction with edge detection that creates thin lines may be problematic if the lines are not very straight, is this a reasonable assumption?

I made an artificial image with a few lines on it.  No matter how I set the angle limits, one line approximately parallel with the y axis was always missed, but it was detected if I checked the Bound option.  This I do not understand at all.  Would anyone have any possible explanations.

I'll go through the tutorial in depth and see if it helps my understanding.
Steven Gentner from United States  [1446 posts] 10 year
Thanks for your notes. There were a couple issues that we resolved with the latest version (just uploaded). Can you download that and try on your test image again?

Note that we use 0 to 180 instead of -90 to 90. Its also cyclical math such that (min to max) 20 to 170 eliminates horiz while 100 to 70 eliminates vertical. We updated the docs to reflect this.

Your assumption on very thin not quite straight lines is correct. That would cause the point to be detected but not be very sharp (thus it could get eliminated if you have too high a sharpness value). That's in part the reason for the Gradient selection which will look at an image gradient (i.e. the step prior to edge detection) which will be responsive for thinner slightly curved lines.

If you continue to have problems isolating particular lines, could you post your original test image here?

Thanks,
STeven.

This forum thread has been closed due to inactivity (more than 4 months) or number of replies (more than 50 messages). Please start a New Post and enter a new forum thread with the appropriate title.

 New Post   Forum Index