imshow, image and imagesc fail on small images

[Antares42]

Hi,

using Octave 3.2.2 with Gnuplot 4.3.0 under Windows, I found the following confusing behaviour:

A=[1 2 3 4; 5 6 7 8];
imshow(A,jet(9));

works while

A=[1 2 3; 5 6 7];
imshow(A,jet(9));

does not. (It produces a white image only.) Similarly,

colormap hot;
A=[1 2 3 4 6; 5 6 7 6 8; 5 5 5 5 6; 5 5 5 5 6; 6 6 6 6 6];
imagesc(A);

works while

colormap hot;
A=[1 2 3 4 6; 5 6 7 6 8; 5 5 5 5 6; 5 5 5 5 6];
imagesc(A);

does not. And last but not least

colormap(gray(2));
image(eye(5)+1);

works, but

colormap(gray(2));
image(eye(4)+1);

does not. (Same of course for image -> imagesc.)

A note: All axes seem to appear correctly. (Or in the case of imshow at least the coordinate system is scaled correctly. Axes can be summoned with "axis on".)

No idea what is going on here.
Daniel

[John W. Eaton-3]

On  3-Nov-2009, Antares42 wrote:

| using Octave 3.2.2 with Gnuplot 4.3.0 under Windows, I found the following
| confusing behaviour:
|
| A=[1 2 3 4; 5 6 7 8];
| imshow(A,jet(9));
|
| works while
|
| A=[1 2 3; 5 6 7];
| imshow(A,jet(9));
|
| does not. (It produces a white image only.) Similarly,
|
| colormap hot;
| A=[1 2 3 4 6; 5 6 7 6 8; 5 5 5 5 6; 5 5 5 5 6; 6 6 6 6 6];
| imagesc(A);
|
| works while
|
| colormap hot;
| A=[1 2 3 4 6; 5 6 7 6 8; 5 5 5 5 6; 5 5 5 5 6];
| imagesc(A);
|
| does not. And last but not least
|
| colormap(gray(2));
| image(eye(5)+1);
|
| works, but
|
| colormap(gray(2));
| image(eye(4)+1);
|
| does not. (Same of course for image -> imagesc.)
|
| A note: All axes seem to appear correctly. (Or in the case of imshow at
| least the coordinate system is scaled correctly. Axes can be summoned with
| "axis on".)

I can't duplicate this problem with Octave 3.2.3 on a Debian system
with gnuplot4.2.6.

jwe
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www-old.cae.wisc.edu/mailman/listinfo/bug-octave

[Antares42]

I tried it on another pc, same configuration, same effect.

Unfortunately, Octave Forge has not yet released 3.2.3 for Windows, so I
cannot test that.

Cheers,
Daniel

John W. Eaton wrote:

> On  3-Nov-2009, Antares42 wrote:
>
> | using Octave 3.2.2 with Gnuplot 4.3.0 under Windows, I found the following
> | confusing behaviour:
> |
> | A=[1 2 3 4; 5 6 7 8];
> | imshow(A,jet(9));
> |
> | works while
> |
> | A=[1 2 3; 5 6 7];
> | imshow(A,jet(9));
> |
> | does not. (It produces a white image only.) Similarly,
> |
> | colormap hot;
> | A=[1 2 3 4 6; 5 6 7 6 8; 5 5 5 5 6; 5 5 5 5 6; 6 6 6 6 6];
> | imagesc(A);
> |
> | works while
> |
> | colormap hot;
> | A=[1 2 3 4 6; 5 6 7 6 8; 5 5 5 5 6; 5 5 5 5 6];
> | imagesc(A);
> |
> | does not. And last but not least
> |
> | colormap(gray(2));
> | image(eye(5)+1);
> |
> | works, but
> |
> | colormap(gray(2));
> | image(eye(4)+1);
> |
> | does not. (Same of course for image -> imagesc.)
> |
> | A note: All axes seem to appear correctly. (Or in the case of imshow at
> | least the coordinate system is scaled correctly. Axes can be summoned with
> | "axis on".)
>
> I can't duplicate this problem with Octave 3.2.3 on a Debian system
> with gnuplot4.2.6.
>
> jwe

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www-old.cae.wisc.edu/mailman/listinfo/bug-octave

[Benjamin Lindner]

> John W. Eaton wrote:
> > On  3-Nov-2009, Antares42 wrote:
> >
> > | using Octave 3.2.2 with Gnuplot 4.3.0 under Windows, I found the
> following
> > | confusing behaviour:
> > |
> > | A=[1 2 3 4; 5 6 7 8];
> > | imshow(A,jet(9));
> > |
> > | works while
> > |
> > | A=[1 2 3; 5 6 7];
> > | imshow(A,jet(9));
> > |
> > | does not. (It produces a white image only.) Similarly,
> > |
> > | colormap hot;
> > | A=[1 2 3 4 6; 5 6 7 6 8; 5 5 5 5 6; 5 5 5 5 6; 6 6 6 6 6];
> > | imagesc(A);
> > |
> > | works while
> > |
> > | colormap hot;
> > | A=[1 2 3 4 6; 5 6 7 6 8; 5 5 5 5 6; 5 5 5 5 6];
> > | imagesc(A);
> > |
> > | does not. And last but not least
> > |
> > | colormap(gray(2));
> > | image(eye(5)+1);
> > |
> > | works, but
> > |
> > | colormap(gray(2));
> > | image(eye(4)+1);
> > |
> > | does not. (Same of course for image -> imagesc.)
> > |
> > | A note: All axes seem to appear correctly. (Or in the case of imshow
> at
> > | least the coordinate system is scaled correctly. Axes can be summoned
> with
> > | "axis on".)
> >
> > I can't duplicate this problem with Octave 3.2.3 on a Debian system
> > with gnuplot4.2.6.
> >
> > jwe
>
> I tried it on another pc, same configuration, same effect.
>
> Unfortunately, Octave Forge has not yet released 3.2.3 for Windows, so I
> cannot test that.

Fortunately, it has (There seem to be problems with SSE3 atlas, so please choose the SSE2 libraries for now)

But you'll find the same problem.
I suspect it's a gnuplot problem with the windows terminal, because if you switch to wxt terminal - putenv("gnuterm","wxt") - the problem's gone.

Debugging octave's commands to gnuplot I find that the commands to set the xrange and yrange are causing the problem. If you omit them (in a separate gnuplot session), then display is fine.

You can get a transcript of the gnuplot commands from octave by

drawnow("windows", "", 0, "debug.gp");

If you then remove the two commands to set xrange and yrange, then it works for me.

A simple testcase that shows the effect is

set term windows
plot "-" matrix with image
1 2
3 1
e
e

versus

set term windows
set xrange [0:1]; set yrange [0:1];
plot "-" matrix with image
1 2
3 1
e
e


This should be reported to gnuplot I guess.

benjamin
benjamin
--
Jetzt kostenlos herunterladen: Internet Explorer 8 und Mozilla Firefox 3.5 -
sicherer, schneller und einfacher! http://portal.gmx.net/de/go/atbrowser
_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www-old.cae.wisc.edu/mailman/listinfo/bug-octave

[Benjamin Lindner]

 >
> This should be reported to gnuplot I guess.
>

I created a thread on gnuplot's mailing list reporting this issue.

benjamin

_______________________________________________
Bug-octave mailing list
Bug-octave@...
https://www-old.cae.wisc.edu/mailman/listinfo/bug-octave