glBeatriz

correcting-image-from-convex-mirror

I cannot see any patch that Jordi mentions when I downloaded the file

I am also very interested in to look at it.

On Jan 16, 2008 9:08 AM, Jordi Puig wrote:

> Wooow your patch is so good! I will love to study it!

> El 16/01/2008, a las 14:34, yair reshef escribió:

https://cycling74.com/forums/index.php?t=msg&goto=117028

> On Jan 16, 2008 2:53 PM, Beatriz < beadelsazp@hotmail.com> wrote:

> > IŽm doing an installation that tracks people walking in a space with

> > As I need the space registered by the camera to be as wide as possible, I

> > My question is, how can I correct the šdistortedš image resulting from the

> mirror to make it šflatšagain? Is there a kind of Quicktime effect for that?

I cannot see any patch that Jordi mentions when I downloaded the file
from this link.
I am also very interested in to look at it.
What did I miss to see?

On Jan 16, 2008 9:08 AM, Jordi Puig  wrote:
>  Wooow your patch is so good! I will love to study it!
>
> j
>
>
> El 16/01/2008, a las 14:34, yair reshef escribió:
>
>
> https://cycling74.com/forums/index.php?t=msg&goto=117028
>
>
> On Jan 16, 2008 2:53 PM, Beatriz < beadelsazp@hotmail.com> wrote:
> >
> >
> > Hi all
> >
> > IŽm doing an installation that tracks people walking in a space with
> cv.jit.objects.
> > As I need the space registered by the camera to be as wide as possible, I
> am using a convex mirror.
> >
> > My question is, how can I correct the šdistortedš image resulting from the
> mirror to make it šflatšagain? Is there a kind of Quicktime effect for that?
> >
> > Thanx!
> > Beatriz
> >
>
>
>


I was checking the patch that Yair posted, it works fine for the specific image which comes with the patch folder, but unfortunately it doesn

it would be easier to help if you could send the image from the mirror.

Make double pass, first with less dots (3X3) to make slight adjustment,

second for more precise (9x13) calculation.

print one dot grid with different color or gray scale value for 3x3 and different for 9x13, with jit.op object select the dots that you need.In that way you are sore that the camera see both grids on the same position.

Make double pass, first with less dots (3X3) to make slight adjustment,
second for more precise (9x13) calculation.

print one  dot grid with different color or gray scale value for 3x3 and different for 9x13, with jit.op object select the dots that you need.In that way you are sore that the camera see both grids on the same position.

Thank you for the good source, At first, ".help" made me confused. ^^

I am wondering if I want to do some projection calibration instead of

camera calibration, can I still apply this code or this kind of

methodolgy for doing that? I am also doing the very similar project

like Beatriz is doing. But I am using a dome mirror instead of convex

> it would be easier to help if you could send the image from the mirror.

> Make double pass, first with less dots (3X3) to make slight adjustment,

> second for more precise (9x13) calculation.

> print one dot grid with different color or gray scale value for 3x3 and different for 9x13, with jit.op object select the dots that you need.In that way you are sore that the camera see both grids on the same position.

Thank you for the good source, At first, ".help" made me confused. ^^
Now I figured out.

I am wondering if I want to do some projection calibration instead of
camera calibration, can I still apply this code or this kind of
methodolgy for doing that? I am also doing the very similar project
like Beatriz is doing. But I am using a dome mirror instead of convex
mirror, so distortion becomes big issue.

On Jan 17, 2008 2:28 PM, jasmin  wrote:
>
> it would be easier to help if you could send the image from the mirror.
>
> just a though
>
> Make double pass, first with less dots (3X3) to make slight adjustment,
> second for more precise (9x13) calculation.
>
> print one  dot grid with different color or gray scale value for 3x3 and different for 9x13, with jit.op object select the dots that you need.In that way you are sore that the camera see both grids on the same position.
>
> Hope i was clear enough.
>
>
>
> thank you Yair for respect
>
>
>
>
> --
> jaz
>


if you want to correct a skewed projection search for [keystone] in the

https://cycling74.com/forums/index.php?t=msg&goto=99242

also, if you want "simple" fisheye correction search for... [fisheye]

there is no silver bullet in that filed, you lose something when distorting

On Jan 17, 2008 11:22 PM, Hyun Jean Lee wrote:

> Thank you for the good source, At first, ".help" made me confused. ^^

> I am wondering if I want to do some projection calibration instead of

> camera calibration, can I still apply this code or this kind of

> methodolgy for doing that? I am also doing the very similar project

> like Beatriz is doing. But I am using a dome mirror instead of convex

> mirror, so distortion becomes big issue.

> > it would be easier to help if you could send the image from the mirror.

> > Make double pass, first with less dots (3X3) to make slight adjustment,

> > second for more precise (9x13) calculation.

> > print one dot grid with different color or gray scale value for 3x3 and

> different for 9x13, with jit.op object select the dots that you need.Inthat way you are sore that the camera see both grids on the same position.

if you want to correct a skewed projection search for [keystone] in the
archives.
see
https://cycling74.com/forums/index.php?t=msg&goto=99242

also, if you want "simple" fisheye correction search for... [fisheye]
there is no silver bullet in that filed, you lose something when distorting
your frame.

On Jan 17, 2008 11:22 PM, Hyun Jean Lee  wrote:

> Hi,
>
> Thank you for the good source, At first, ".help" made me confused. ^^
> Now I figured out.
>
> I am wondering if I want to do some projection calibration instead of
> camera calibration, can I still apply this code or this kind of
> methodolgy for doing that? I am also doing the very similar project
> like Beatriz is doing. But I am using a dome mirror instead of convex
> mirror, so distortion becomes big issue.
>
> Thanks,
>
>
> Hyun
>
>
>
>
> On Jan 17, 2008 2:28 PM, jasmin  wrote:
> >
> > it would be easier to help if you could send the image from the mirror.
> >
> > just a though
> >
> > Make double pass, first with less dots (3X3) to make slight adjustment,
> > second for more precise (9x13) calculation.
> >
> > print one  dot grid with different color or gray scale value for 3x3 and
> different for 9x13, with jit.op object select the dots that you need.Inthat way you are sore that the camera see both grids on the same position.
> >
> > Hope i was clear enough.
> >
> >
> >
> > thank you Yair for respect
> >
> >
> >
> >
> > --
> > jaz
> >
>

Btw, it probably isn't as general, but if you just want simple lens 

correction, check the archive link below. Something along these lines 

could also work for a mirror when shooting from the right angle.

https://cycling74.com/forums/index.php?t=msg&goto=88305l#msg_88305

Btw, it probably isn't as general, but if you just want simple lens  
correction, check the archive link below. Something along these lines  
could also work for a mirror when shooting from the right angle.

Sooooo noow I start to understand the patch. The grid is the "reference" for jit.repos to modify the image...you people are my heroes! ;)

Thanks Jazmataz and Criccrec, and thanks Yari for your clear explanation! I will start to experiment....

Correcting image from convex mirror