Re: imwrite.m test failures

[marco atzeri]

Il 10/7/2013 11:38 AM, Andreas Weber ha scritto:
> On 06.10.2013 19:03, marco atzeri wrote:
> > Il 10/6/2013 6:52 PM, Andreas Weber ha scritto:
> > > Am 03.10.2013 21:46, schrieb marco atzeri:
> > > > testing latest devel
> > > > (changeset:   17540:b7aa1d726bcc)
> > > >
> > > > anyone seeing the same ?
> > > >
> > > > > > > > > processing /pub/hg/octave/scripts/image/imwrite.m
> > > >     ***** testif HAVE_MAGICK
> > > >    imw = randi (255, 100, "uint8");
> > > >    filename = [tmpnam() ".png"];
> > > >    unwind_protect
> > > >      imwrite (imw, filename);
> > > >      imr = imread (filename);
> > > >    unwind_protect_cleanup
> > > >      unlink (filename);
> > > >    end_unwind_protect
> > > >    assert (imw, imr)
> > > > !!!!! test failed
> > > > ASSERT errors for:  assert (imw,imr)
> > > >
> > > >     Location  |  Observed  |  Expected  |  Reason
> > > >      (1,1)         104          103        Abs err 1 exceeds tol 0
> > > >      (2,1)         184          183        Abs err 1 exceeds tol 0
> > > > ..
> > > >     (99,100)        96           95        Abs err 1 exceeds tol 0
> > > >    (100,100)       126          125        Abs err 1 exceeds tol 0
> > > > > > > > > processing /pub/hg/octave/scripts/image/ind2gray.m
> > >
> > > Hi Marco,
> > > cset: 0c6750ace8b4, test imwrite.m PASSES 10 out of 10 tests
> > >
> > > GraphicsMagick 1.3.16 2012-06-23 Q8 http://www.GraphicsMagick.org/
> > >
> > > MAGICK_CPPFLAGS = -I/usr/include/GraphicsMagick
> > > MAGICK_LDFLAGS =
> > > MAGICK_LIBS = -lGraphicsMagick++ -lGraphicsMagick
> > >
> > > Can you try
> > > octave> imw = uint8([0 100 200]);
> > > octave> imwrite(imw, "testA.png")
> > >
> > > and then in a shell
> > > $ gm identify -verbose testA.png
> > >
> > > I get:
> > > Image: testA.png
> > >     Format: PNG (Portable Network Graphics)
> > >     Geometry: 3x1
> > >     Class: PseudoClass
> > >     Type: grayscale
> > >     Depth: 8 bits-per-pixel component
> > >     Channel Depths:
> > >       Gray:     6 bits
> > >     Channel Statistics:
> > >       Gray:
> > >         Minimum:                     0.00 (0.0000)
> > >         Maximum:                   200.00 (0.7843)
> > >         Mean:                      100.00 (0.3922)
> > >         Standard Deviation:        100.00 (0.3922)
> > >     Colors: 256
> > >       0: (  0,  0,  0)      black
> > >       1: (  1,  1,  1)      #010101
> > > ..
> > >       100: (100,100,100)      #646464
> > > ...
> > >       200: (200,200,200)      #C8C8C8
> > > ...
> > >     255: (255,255,255)      white
> > >     Filesize: 69
> > >     Interlace: No
> > >     Orientation: Unknown
> > >     Background Color: white
> > >     Border Color: #DFDFDF
> > >     Matte Color: #BDBDBD
> > >     Compose: Over
> > >     Dispose: Undefined
> > >     Iterations: 0
> > >     Compression: Zip
> > >     Signature:
> > > 6f5c5c4f22b887737ef219b11c1e7f94faef02e50d985032538327b2bcd9a2ec
> > >     Tainted: False
> > >
> > > Until now it could be a problem with imwrite or imread, perhaps we can
> > > assign it to one of them.
> > >
> > > HTH, Andy
> >
> > Q16 instead of Q8
> > GraphicsMagick 1.3.18 2013-03-10 Q16 http://www.GraphicsMagick.org/
> >
> > therea are some slight differences in the gray depth
> >
> > $ gm identify -verbose testA.png
> > Image: testA.png
> >    Format: PNG (Portable Network Graphics)
> >    Geometry: 3x1
> >    Class: PseudoClass
> >    Type: grayscale
> >    Depth: 8 bits-per-pixel component
> >    Channel Depths:
> >      Gray:     8 bits
> >    Channel Statistics:
> >      Gray:
> >        Minimum:                     0.00 (0.0000)
> >        Maximum:                 51143.00 (0.7804)
> >        Mean:                    25528.67 (0.3895)
> >        Standard Deviation:      25571.61 (0.3902)
> >    Colors: 256
> >      0: (  0,  0,  0)      black
> >      1: (  1,  1,  1)      #010101
> >
> >      100: (100,100,100)    #646464
> >
> >      255: (255,255,255)    white
> >    Filesize: 69
> >    Interlace: No
> >    Orientation: Unknown
> >    Background Color: white
> >    Border Color: #DFDFDF
> >    Matte Color: #BDBDBD
> >    Compose: Over
> >    Dispose: Undefined
> >    Iterations: 0
> >    Compression: Zip
> >    Png:IHDR.color-type-orig: 0
> >    Png:IHDR.bit-depth-orig: 8
> >    Signature:
> > eaa4e546aa78466f45e563dbcbab6f26d66e96f2485b3e199d7b0e59c5f49775
> >    Tainted: False
>
> Hi Marco,
> I use octave default 36b9fa789d8e
> can you try this changeset too? There were some changes to imwrite since
> b7aa1d726bcc.
>
> I compiled libgraphicsmagick with Q16:
> GraphicsMagick 1.3.16 2012-06-23 Q16 http://www.GraphicsMagick.org/
>
> $ gm identify -verbose testA.png
> Image: testA.png
>    Format: PNG (Portable Network Graphics)
>    Geometry: 3x1
>    Class: PseudoClass
>    Type: grayscale
>    Depth: 8 bits-per-pixel component
>    Channel Depths:
>      Gray:     8 bits
>    Channel Statistics:
>      Gray:
>        Minimum:                     0.00 (0.0000)
>        Maximum:                 51400.00 (0.7843)
>        Mean:                    25700.00 (0.3922)
>        Standard Deviation:      25700.00 (0.3922)
>
>
> Do you see the different values for Max and Mean?


$ gm identify -verbose testA1.png
Image: testA1.png
  Format: PNG (Portable Network Graphics)
  Geometry: 3x1
  Class: PseudoClass
  Type: grayscale
  Depth: 8 bits-per-pixel component
  Channel Depths:
    Gray:     8 bits
  Channel Statistics:
    Gray:
      Minimum:                     0.00 (0.0000)
      Maximum:                 51143.00 (0.7804)
      Mean:                    25528.67 (0.3895)
      Standard Deviation:      25571.61 (0.3902)


> When I read back the image with imread I get
> octave:1> imread("testA.png")
> warning: your version of GraphicsMagick limits images to 16 bits per pixel
> ans =
>
>      0  100  200
>
> So perhaps a problem with the newer gm lib?
>
> Regards Andy

 imread("testA1.png")
ans =

    0   99  199

I guess it is a rounding issue between different platforms

Regards
Marco