Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Some Questions about Hough Transforms...

4 views
Skip to first unread message

SpreadTooThin

unread,
Apr 16, 2008, 7:26:28 PM4/16/08
to
How can I determine the length of the line that the hough transform
detects?

What goes on with aliased lines...
for example..

X
X
X
X
X
X
X


Is this three vertical lines or is it one line?

I seem to remember there being a modified hough transform that would
tell you were the line started and ended.. anyone else seen that
paper?
If so.. Got a link?

lomas

unread,
Apr 16, 2008, 9:02:25 PM4/16/08
to

In my opinion,
whether they are one line or not is depend on the bin you use in hough
transform parameter space.
and after hough transfrom, the bin where peak lying holds the
length(in pixel) and line parameters (e.g. K and B in y = K*x+ B).
With these parameters, the ending of line is possible to be found by
some discussion.

Best Regards,
lomas

JS

unread,
Apr 18, 2008, 7:12:02 AM4/18/08
to
> Is this three vertical lines or is it one line?
That will depend on the resolution of your accumulator.

Determining the length of your line: if the input to the hough
transform is a binary image, then the length of any possible line
could be considered to be the value in the corresponding accumulator
bin. However, I would not recommend using this as the length. I
think that the HT is good for getting general positions of lines, but
if you want length/position of lines you need to then use other
methods as well.

SpreadTooThin

unread,
Apr 18, 2008, 5:57:45 PM4/18/08
to
On Apr 18, 5:12 am, JS <jnospams...@gmail.com> wrote:
> > Is this three vertical lines or is it one line?
>
> That will depend on the resolution of your accumulator.
>
And this is really the problem I'm facing.. How do I determine the
size of the accumulator?
I can find the maximum peak easily, but what about other peaks in the
image...
How can I go about finding those?

I'm wondering if I should find a peak and then use the neighbors as a
weighted sum to get
the actual R Theta... Make sence?

Nicolas Bonneel

unread,
Apr 19, 2008, 7:34:09 AM4/19/08
to
SpreadTooThin a écrit :

You can increase the dimension of your Hough transform to make it 4
dimensions instead of 2 (for example the 4 dimensions are the
x1,y1,x2,y2) and find your maximum in this 4 dimensional space.

You will directly get your x1,y1,x2,y2 and the corresponding length.

Otherwise, I don't really know how to get the length in the usual Hough
transform since it only encodes directions.


--
Nicolas Bonneel
http://www.bonneel.com

Norberto Goussies

unread,
May 6, 2008, 9:18:49 AM5/6/08
to
> How can I determine the length of the line that the hough transform
> detects?

MATLAB has the houghlines function to get the length. I have tried it
is very useful. You must provide two parameters : 'FillGap' and
'MinLength'

SpreadTooThin

unread,
May 6, 2008, 11:13:24 PM5/6/08
to

Oddly enough thats exactly what I did in my application.
Cool.

0 new messages