From dkirtley at gmail.com  Wed May  4 18:35:25 2005
From: dkirtley at gmail.com (David Kirtley)
Date: Wed, 4 May 2005 11:35:25 -0500
Subject: [Image-SIG] Odd behavior in PIL 1.1.4 equalization
Message-ID: <22b0f40d050504093530609811@mail.gmail.com>

Hi all, 

I seem to have found an odd behavior in the equalization.

I created a series of test images with PIL for a demonstration:
  a 300x300 RGB image with three bars of gray with offsets of 20 in
all three channels
 with starting values of 0 to 200 in steps of 50.

I expected the equalization routine to move them to 0, 128, 255 as per
the definition of Histogram Equailzation but to my suprise the
equalization routine decides to instead
set them to 0, 85, 170.

Is this intended functionality? Is this a bug/feature?  :)


David

------------------------------------- Source used-------------------------


import Image, ImageOps, ImageDraw, ImageFilter

def fill(im,start):
    d = ImageDraw.Draw(im)
    
    d.rectangle((0,0,99,299), fill=(start+0,start+0,start+0), outline
=(start+0,start+0,start+0))
    d.rectangle((100,0,199,299), fill=(start+20,start+20,start+20),
outline =(start+20,start+20,start+20))
    d.rectangle((200,0,299,299), fill=(start+40,start+40,start+40),
outline =(start+40,start+40,start+40))

    return im

x = Image.new('RGB',(300,300))

original = open('original.csv','w')
equalized = open('equalized.csv','w')
for image_base in [a for a in range(0,200,50)]: 
    a = fill(x, image_base)
    a.save('%s_original.png'%image_base,'PPM')
    b  = ImageOps.equalize(a) 
    b.save('%s_equalized.png'%image_base,'PPM')

    original.write('%s,'%image_base+`a.histogram()`[1:-1]+'\n')
    equalized.write('%s,'%image_base+`b.histogram()`[1:-1]+'\n')

From darkstar at brturbo.com.br  Wed May  4 22:50:17 2005
From: darkstar at brturbo.com.br (Weldner Alves)
Date: Wed, 4 May 2005 17:50:17 -0300 (BRT)
Subject: [Image-SIG] Question about PIL and transparency
Message-ID: <31227331.1115239817089.JavaMail.nobody@webmail1.brturbo.com>

What I wanna do is simple:
I just want to write a string on a transparent background. Here is my code:

import Image, ImageDraw, ImageFont

im = Image.new("P",(500,500), 'white')
fonte = ImageFont.truetype("comicbd.ttf",15)
draw = ImageDraw.Draw(im)

im.info['transparency'] = 255 
"""(the background is white, then white is my transparency)"""
draw.text((0,0), "HERE GOES THE STRING", fill='black', font=fonte)

#save
im.save("something.png", "PNG")


Now when I go to the prompt:

>>> im = Image.open('something.png')
>>> im.info.items()
[]
>>> im.info["transparency"]=255
>>> im.save("anything.png", "PNG")
>>> im = Image.open('anything.png')
>>> im.info.items()
[]

There was suposed to be a "transparency" with value 255 inside the info dict.
So whats wrong ?

From fredrik at pythonware.com  Thu May  5 11:05:43 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Thu, 5 May 2005 11:05:43 +0200
Subject: [Image-SIG] Question about PIL and transparency
References: <31227331.1115239817089.JavaMail.nobody@webmail1.brturbo.com>
Message-ID: <d5cnob$ds4$3@sea.gmane.org>

Weldner Alves wrote:

> im.info['transparency'] = 255
> #save
> im.save("something.png", "PNG")

try

    im.save("something.png", transparency=255)

instead.

</F>




From fredrik at pythonware.com  Thu May  5 11:04:31 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Thu, 5 May 2005 11:04:31 +0200
Subject: [Image-SIG] Odd behavior in PIL 1.1.4 equalization
References: <22b0f40d050504093530609811@mail.gmail.com>
Message-ID: <d5cnob$ds4$2@sea.gmane.org>

David Kirtley wrote:

> I seem to have found an odd behavior in the equalization.
>
> I created a series of test images with PIL for a demonstration:
>   a 300x300 RGB image with three bars of gray with offsets of 20 in
> all three channels
>  with starting values of 0 to 200 in steps of 50.
>
> I expected the equalization routine to move them to 0, 128, 255 as per
> the definition of Histogram Equailzation but to my suprise the
> equalization routine decides to instead
> set them to 0, 85, 170.
>
> Is this intended functionality? Is this a bug/feature?  :)

looks like an off-by-one error (note that 170+85=255).  should only
cause problems for extreme cases, and should be trivial to fix.

I'll into this for 1.1.6 (patches are welcome).

</F>




From fredrik at pythonware.com  Thu May  5 23:14:10 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Thu, 5 May 2005 23:14:10 +0200
Subject: [Image-SIG] Odd behavior in PIL 1.1.4 equalization
References: <22b0f40d050504093530609811@mail.gmail.com>
	<d5cnob$ds4$2@sea.gmane.org>
Message-ID: <d5e44v$jg6$2@sea.gmane.org>

I wrote:

> I'll into this for 1.1.6

make that "I'll look into this for 1.1.6".

> (patches are welcome).

I've attached a (preliminary) patch.

</F>

Index: PIL/ImageOps.py
===================================================================
--- PIL/ImageOps.py     (revision 2382)
+++ PIL/ImageOps.py     (working copy)
@@ -207,13 +207,15 @@
     h = image.histogram(mask)
     lut = []
     for b in range(0, len(h), 256):
-        step = reduce(operator.add, h[b:b+256]) / 255
-        if step == 0:
-            step = 1
-        n = 0
-        for i in range(256):
-            lut.append(n / step)
-            n = n + h[i+b]
+        histo = filter(None, h[b:b+256])
+        if len(histo) <= 1:
+            lut.append(range(256))
+        else:
+            step = (reduce(operator.add, histo) - histo[-1]) / 255
+            n = step / 2
+            for i in range(256):
+                lut.append(n / step)
+                n = n + h[i+b]
     return _lut(image, lut)

 ##




From douglas at paradise.net.nz  Fri May  6 04:00:04 2005
From: douglas at paradise.net.nz (Douglas Bagnall)
Date: Fri, 06 May 2005 14:00:04 +1200
Subject: [Image-SIG] downscaling by area averaging
Message-ID: <427ACFA4.6070602@paradise.net.nz>


hi,

I've made a patch to resize by averaging pixels.

http://halo.gen.nz/pil/pil_area_average.patch

If people want it, I'll fix up the GCCisms, and make it work on other 
than RGB images.

I wrote it for statistical analysis of images (to get the average 
density of various areas).

For every size I tried it proved quicker and more accurate than the 
antialias filter (which, OTOH, is more flexible).

Typical times resizing from 450 x 325 to 16 x 12 are:
         NEAREST:             0.000022
        BILINEAR:             0.000112
         BICUBIC:             0.000192
    AREA_AVERAGE:             0.001558
       ANTIALIAS:             0.009321

(linux, athlon XP 2800+).

The results of this are here:
http://halo.gen.nz/pil/examples.html

Interestingly, the area averaged image is generally pixel perfect with 
the Gimp's best resizing ("Cubic").

and finally, I just discovered that emacs can save files as syntax 
highlighted html, so here is the guts of the patch in full colour:

http://halo.gen.nz/pil/AreaAverage_c.html


regards

Douglas Bagnall

From fredrik at pythonware.com  Fri May  6 12:19:32 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Fri, 6 May 2005 12:19:32 +0200
Subject: [Image-SIG] New to PIL and struggling...
References: <opspm1a3vyyt02yl@mail.isis.de>
Message-ID: <d5ffus$9td$1@sea.gmane.org>

"charlie clark" wrote:

> Trying to understand things: lut is effectively a palette table all set to
> the same value?
> What does lut[im.info["transparency"]] = 0 do?
> In my case its would sets lut[14] = 0
> So the real magic is im.point() which creates a single band presumably
> reacting to the transparency in the original image.

"point" just maps the source pixels through the lookup-table
(the "lut"), like this:

    for all coordinates in image:
        ... load pixel from input image ...
        pixel = lut[pixel]
        ... store pixel in output pixel ...

the result is a mask which has the value 0 for all transparent pixels
(index 14), and the value 1 for all other pixels.  "point" itself doesn't
know about transparency; it just uses the table it got.

</F>




From fredrik at pythonware.com  Fri May  6 12:26:34 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Fri, 6 May 2005 12:26:34 +0200
Subject: [Image-SIG] downscaling by area averaging
References: <427ACFA4.6070602@paradise.net.nz>
Message-ID: <d5fgc3$b50$1@sea.gmane.org>

Douglas Bagnall wrote:

> I've made a patch to resize by averaging pixels.
>
> http://halo.gen.nz/pil/pil_area_average.patch
>
> If people want it, I'll fix up the GCCisms, and make it work on other
> than RGB images.

looks interesting, but in theory (and without analyzing your
code), shouldn't this be the same as running ImagingStretch
with the nearest filter?  (=an adaptive box filter)

(guess I'll have to do some experiments...  I've been meaning to
switch to ImagingStretch for all downsampling uses, but haven't
had time to tune the filters yet (=make them behave the same
as the upsampling filters).  this might be the prodding I needed)

> Interestingly, the area averaged image is generally pixel perfect with
> the Gimp's best resizing ("Cubic").

that's odd.  what kind of source images are you processing?

</F>




From fredrik at pythonware.com  Fri May  6 13:29:58 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Fri, 6 May 2005 13:29:58 +0200
Subject: [Image-SIG] downscaling by area averaging
References: <427ACFA4.6070602@paradise.net.nz> <d5fgc3$b50$1@sea.gmane.org>
Message-ID: <d5fk30$n30$1@sea.gmane.org>

> what kind of source images are you processing?

this is what happens when you don't click on all links in a
message ;-)

some observations:

+ the sample you used sure illustrates why PIL's BILINEAR and
  BICUBIC fixed-size kernel filters are not very good for down-
  sampling...

+ in case anyone wants to play with PIL's internal stretch filters,
  you can use the following little wrapper function:

    def stretch(im, size, filter=Image.NEAREST):
        im.load()
        im = im._new(im.im.stretch(size, filter))
        return im

  supported filters are Image.NEAREST, BILINEAR, and BICUBIC,
  in addition to ANTIALIAS (which is already exposed via the
  "resize" method).  unlike their resize counterparts, the stretch
  filters use an adaptive kernel size (the more you downsample,
  the larger the kernel)

+ as expected, NEAREST stretch isn't too far from your results, but
  it's not identical (and it probably shouldn't be)

+ I'm beginning to wonder if there's not an off-by-something error
  lurking somewhere in stretch.  hmm...

</F>




From darkstar at brturbo.com.br  Fri May  6 19:33:45 2005
From: darkstar at brturbo.com.br (Weldner Alves)
Date: Fri, 6 May 2005 14:33:45 -0300 (BRT)
Subject: [Image-SIG] Question about PIL and transparency
Message-ID: <31780687.1115400825585.JavaMail.nobody@webmail4.brturbo.com>

What I wanna do is simple:
I just want to write a string on a transparent background. Here is my code:

import Image, ImageDraw, ImageFont

im = Image.new("P",(500,500), 'white')
fonte = ImageFont.truetype("comicbd.ttf",15)
draw = ImageDraw.Draw(im)

im.info['transparency'] = 255 
"""(the background is white, then white is my transparency)"""
draw.text((0,0), "HERE GOES THE STRING", fill='black', font=fonte)

#save
im.save("something.png", "PNG")


Now when I go to the prompt:

>>> im = Image.open('something.png')
>>> im.info.items()
[]
>>> im.info["transparency"]=255
>>> im.save("anything.png", "PNG")
>>> im = Image.open('anything.png')
>>> im.info.items()
[]

There was suposed to be a "transparency" with value 255 inside the info dict.
So whats wrong ?

From fredrik at pythonware.com  Fri May  6 20:06:35 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Fri, 6 May 2005 20:06:35 +0200
Subject: [Image-SIG] Question about PIL and transparency
References: <31780687.1115400825585.JavaMail.nobody@webmail4.brturbo.com>
Message-ID: <d5gbal$gpv$1@sea.gmane.org>

Weldner Alves wrote:

> What I wanna do is simple:
> I just want to write a string on a transparent background. Here is my code:

did you post this again after I replied, or is it a problem with the
python.org mailserver?  (if the former, please read replies before
repeating a question).

</F>




From douglas at paradise.net.nz  Sat May  7 03:54:28 2005
From: douglas at paradise.net.nz (Douglas Bagnall)
Date: Sat, 07 May 2005 13:54:28 +1200
Subject: [Image-SIG] downscaling by area averaging
In-Reply-To: <d5fk30$n30$1@sea.gmane.org>
References: <427ACFA4.6070602@paradise.net.nz> <d5fgc3$b50$1@sea.gmane.org>
	<d5fk30$n30$1@sea.gmane.org>
Message-ID: <427C1FD4.70700@paradise.net.nz>


I've added stretched versions to

http://halo.gen.nz/pil/examples.html

With javascript, clicking on the top image should cycle through the
unscaled, area averaged, anti-aliased, and nearest-neighbour-stetched
images.

> + I'm beginning to wonder if there's not an off-by-something error
>   lurking somewhere in stretch.  hmm...

In my tests, all the resize and stretch filters appear to shift the 
image down and to the right by about half an output pixel.  Assuming my 
controls images are right...


[Earlier]

 >>Interestingly, the area averaged image is generally pixel perfect with
 >>the Gimp's best resizing ("Cubic").
 >
 > that's odd.

Without looking, I think it must be adaptive, but is called cubic 
because that is an ancient photoshop keyword for "the best possible".


douglas


From fredrik at pythonware.com  Sat May  7 11:09:14 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Sat, 7 May 2005 11:09:14 +0200
Subject: [Image-SIG] Odd behavior in PIL 1.1.4 equalization
References: <22b0f40d050504093530609811@mail.gmail.com><d5cnob$ds4$2@sea.gmane.org>
	<d5e44v$jg6$2@sea.gmane.org>
Message-ID: <d5i06r$epk$1@sea.gmane.org>

> I've attached a (preliminary) patch.

here's a sligtly less broken patch:

Index: PIL/ImageOps.py
===================================================================
--- PIL/ImageOps.py     (revision 2134)
+++ PIL/ImageOps.py     (working copy)
@@ -9,6 +9,7 @@
 # 2001-10-23 fl   Added autocontrast operator
 # 2001-12-18 fl   Added Kevin's fit operator
 # 2004-03-14 fl   Fixed potential division by zero in equalize
+# 2005-05-05 fl   Fixed equalize for low number of values
 #
 # Copyright (c) 2001-2004 by Secret Labs AB
 # Copyright (c) 2001-2004 by Fredrik Lundh
@@ -207,13 +208,18 @@
     h = image.histogram(mask)
     lut = []
     for b in range(0, len(h), 256):
-        step = reduce(operator.add, h[b:b+256]) / 255
-        if step == 0:
-            step = 1
-        n = 0
-        for i in range(256):
-            lut.append(n / step)
-            n = n + h[i+b]
+        histo = filter(None, h[b:b+256])
+        if len(histo) <= 1:
+            lut.extend(range(256))
+        else:
+            step = (reduce(operator.add, histo) - histo[-1]) / 255
+            if not step:
+                lut.extend(range(256))
+            else:
+                n = step / 2
+                for i in range(256):
+                    lut.append(n / step)
+                    n = n + h[i+b]
     return _lut(image, lut)

 ##




From jelle.feringa at ezct.net  Sat May  7 12:27:18 2005
From: jelle.feringa at ezct.net (Jelle Feringa / EZCT Architecture & Design Research)
Date: Sat, 7 May 2005 12:27:18 +0200
Subject: [Image-SIG] loading PIL image to pygame without the need to write a
	temporary image file
Message-ID: <20050507102404.073071E4003@bag.python.org>

Hi,

I'm working on a script where I blend frames according to their visual
difference.
So if there's not too much difference, you'll have a swift fade-in, if the
difference between the two frames is greater, the fade-in will last longer.
Simple idea, but very effective, so cool to be able to do that kind of thing
< 15 minutes in PIL & python.

I'm using this sequence to write directly to mpeg2 output, and that's where
things mess up.
I use PyMedia (worth checking out!) to encode to mpeg2 files.
My guess was that pilmage.tostring() would be quite similar to pygame's 
pygame.image.tostring(img, "RGB") function. That unfortuanately isn't the
case. 

When I use the PIL function to encode to a string, the program crashes after
a couple of frames, and works fine with the pygame string function. Since I
use a PIL blend method to get the magic done, I'm trying to find a way to
load the image into pygame without needing to write a temporary image file.

And that's where I'm left clueless. 
I'm fairly new to python, and this is the kind of thing that still gets me
stuck.
So I hope to be able to learn from someone's experience facing a similar
problem.

Cheers,

Jelle


From fredrik at pythonware.com  Sat May  7 12:29:07 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Sat, 7 May 2005 12:29:07 +0200
Subject: [Image-SIG] downscaling by area averaging
References: <427ACFA4.6070602@paradise.net.nz>
	<d5fgc3$b50$1@sea.gmane.org><d5fk30$n30$1@sea.gmane.org>
	<427C1FD4.70700@paradise.net.nz>
Message-ID: <d5i4sl$pbq$1@sea.gmane.org>

Douglas Bagnall wrote:

> > + I'm beginning to wonder if there's not an off-by-something error
> >   lurking somewhere in stretch.  hmm...
>
> In my tests, all the resize and stretch filters appear to shift the
> image down and to the right by about half an output pixel.  Assuming
> my controls images are right...

I've attached a (preliminary) patch.

</F>

Index: libImaging/Antialias.c
===================================================================
--- libImaging/Antialias.c      (revision 2394)
+++ libImaging/Antialias.c      (working copy)
@@ -145,7 +145,7 @@
     if (imIn->xsize == imOut->xsize) {
         /* vertical stretch */
         for (yy = 0; yy < imOut->ysize; yy++) {
-            center = yy * scale;
+            center = (yy + 0.5) * scale;
             ww = 0.0;
             ss = 1.0 / filterscale;
             /* calculate filter weights */
@@ -222,7 +222,7 @@
     } else {
         /* horizontal stretch */
         for (xx = 0; xx < imOut->xsize; xx++) {
-            center = xx * scale;
+            center = (xx + 0.5) * scale;
             ww = 0.0;
             ss = 1.0 / filterscale;
             xmin = floor(center - support);




From fredrik at pythonware.com  Sat May  7 14:41:01 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Sat, 7 May 2005 14:41:01 +0200
Subject: [Image-SIG] loading PIL image to pygame without the need to
	write atemporary image file
References: <20050507102404.073071E4003@bag.python.org>
Message-ID: <d5icjs$c7r$1@sea.gmane.org>

Jelle Feringa wrote:

> My guess was that pilmage.tostring() would be quite similar to pygame's
> pygame.image.tostring(img, "RGB") function. That unfortuanately isn't the
> case.

judging from the pygame documentation, they do the exact same thing
(but with slightly different arguments).  according to the documentation,
the following should work:

    mode = image.mode
    size = image.size
    data = image.tostring()

    assert mode in "RGB", "RGBA"

    surface = pygame.image.fromstring(data, size, mode)

(and you could probably use frombuffer instead of fromstring).

> When I use the PIL function to encode to a string, the program crashes after
> a couple of frames, and works fine with the pygame string function.

it crashes and works fine?  how does it crash?  traceback?  coredump?
what crashes and what works?

</F>




From jelle.feringa at ezct.net  Sat May  7 16:25:16 2005
From: jelle.feringa at ezct.net (Jelle Feringa / EZCT Architecture & Design Research)
Date: Sat, 7 May 2005 16:25:16 +0200
Subject: [Image-SIG]  loading PIL image to pygame without the need to
Message-ID: <20050507142109.4BC691E4003@bag.python.org>

Hi Frederik,

That definitely solved my troubles, thanks a lot!
Sorry for the crash confusion; PyMedia so far has no polite way of saying
'the data you supplied is corrupt' but grinds python to a halt.
Is use the following code now, which is working just swell!

Cheers,

Jelle.

            img = pygame.image.frombuffer(blend.tostring(), blend.size,
blend.mode)
            img = pygame.image.tostring(img, "RGB")
            bmpFrame= vcodec.VFrame(vcodec.formats.PIX_FMT_RGB24, sizeBlend,
(blend.tostring(),None,None))
            print 'bmpFrame'
            yuvFrame= bmpFrame.convert(vcodec.formats.PIX_FMT_YUV420P)
            print 'yuvFrame'
            d = e.encode(yuvFrame)
            print 'encode yuv - frame'
            video.write(d)
            print 'write frame to video file'


From m.thomas at ucl.ac.uk  Sun May  8 18:48:31 2005
From: m.thomas at ucl.ac.uk (Mark Thomas)
Date: Sun, 08 May 2005 17:48:31 +0100
Subject: [Image-SIG] Help with PIL - importing CGM files.
Message-ID: <BEA4016F.893C%m.thomas@ucl.ac.uk>

Does anybody know how I can import / load CGM image files into the Python
Imaging Library?
It seems odd to me that this isn't one of the standard recognized formats
for PIL since it is the main output file format for a number of python
graphics modules (gist, oog etc).

Thanks

Mark
-- 
Dr Mark Thomas
Department of Biology
University College London
Gower Street
London WC1E 6BT
Tel:   (020) 7 679 2654
Fax:   (020) 7 679 7096
Mobile: 07973 725955
Email: m.thomas at ucl.ac.uk
Web:   www.ucl.ac.uk/tcga/



From lbolognini at gmail.com  Tue May 10 01:31:12 2005
From: lbolognini at gmail.com (Lorenzo Bolognini)
Date: Tue, 10 May 2005 00:31:12 +0100
Subject: [Image-SIG] CAPTCHA antibot
Message-ID: <a654ded205050916315e4388c8@mail.gmail.com>

Hi all,

first post here ;)

i'm trying to do this:
from PIL import Image, ImageDraw, ImageFont
im = Image.new("RGB", (150,70), "#fff")
draw = ImageDraw.Draw(im, "RGB")
font = ImageFont.truetype("trebuc.ttf", 30)
draw.text((20,20), "Hello World!", fill="#ff0000", font=font)
im.save("C:\\pippo.png", "PNG")

Trying to achieve something like this:
http://www.google.com/accounts/

Google doesn't have a complex antibot protection but it's cute and
it's good enough for me.

What I'm missing in the code above is the deformation which the PIL
documentation says it's supposed to be achieved by instatiating a
deform object. I searched this ML archives but the only question on
this subject got unanswered.

So the question is: how do I create a deform object?

Thanks everybody,
Lorenzo

From douglas at paradise.net.nz  Tue May 10 06:07:36 2005
From: douglas at paradise.net.nz (Douglas Bagnall)
Date: Tue, 10 May 2005 16:07:36 +1200
Subject: [Image-SIG] CAPTCHA antibot
In-Reply-To: <a654ded205050916315e4388c8@mail.gmail.com>
References: <a654ded205050916315e4388c8@mail.gmail.com>
Message-ID: <42803388.50308@paradise.net.nz>

Lorenzo Bolognini wrote:

> from PIL import Image, ImageDraw, ImageFont
> im = Image.new("RGB", (150,70), "#fff")
> draw = ImageDraw.Draw(im, "RGB")
> font = ImageFont.truetype("trebuc.ttf", 30)
> draw.text((20,20), "Hello World!", fill="#ff0000", font=font)
> im.save("C:\\pippo.png", "PNG")

[...]
> So the question is: how do I create a deform object?

ImageOps.deform is precisely this:

def deform(image, deformer, resample=Image.BILINEAR):
     "Deform image using the given deformer"
     return image.transform(
         image.size, Image.MESH, deformer.getmesh(image), resample
         )

so a deformer object is something with a getmesh method that returns a
deformation mesh in response to an image.  Your deformer object could do
  something like find the pixel length of the word and return the
appropriate filter.  But you probably don't need that.  A simple quad
filter might be enough:

# corner co-ordinates, anticlockwise from top left
corners = (15,15, 18,60, 140,50, 130,10)
im.transform(im.size, Image.QUAD, corners,
              Image.BICUBIC).save("/tmp/pippo2.png")

> Trying to achieve something like this:
> http://www.google.com/accounts/

I think you mean https://www.google.com/accounts/NewAccount


douglas



From douglas at paradise.net.nz  Fri May 13 03:56:12 2005
From: douglas at paradise.net.nz (Douglas Bagnall)
Date: Fri, 13 May 2005 13:56:12 +1200
Subject: [Image-SIG] another out by 0.5 patch
In-Reply-To: <d5i4sl$pbq$1@sea.gmane.org>
References: <427ACFA4.6070602@paradise.net.nz> <d5fgc3$b50$1@sea.gmane.org>
	<d5fk30$n30$1@sea.gmane.org> <427C1FD4.70700@paradise.net.nz>
	<d5i4sl$pbq$1@sea.gmane.org>
Message-ID: <4284093C.10104@paradise.net.nz>

Fredrik Lundh wrote:

>>In my tests, all the resize and stretch filters appear to shift the
>>image down and to the right by about half an output pixel. 
>  
> I've attached a (preliminary) patch.
> 

It certainly works for me.  Area-averaged and stretched images are 
mostly visually indistinguishable -- the exceptions being high contrast 
drawings.

Stretch-resized images are consistently 1/256th darker than the same 
images resized by other means.  My understanding of this is that when
a region of input pixels is (say) all 255s, except for one pixel of 254, 
then the float average is 254.9xx, so the output pixel becomes 254.  On 
average this should darken by 0.5, but over the two passes, it darkens 
by a unit.  I've attached a patch.


Here's an example, resizing a 450x325 image to 16x12 (as seen on 
http://halo.gen.nz/pil/examples.html), showing mean, RMS, and extreme 
differences between pixels.

before the patch, comparing stretched(NEAREST) to gimp:

R:  mean: 1.020833  RMS: 1.190238, extrema: (-2.0, 4.0)
G:  mean: 0.973958  RMS: 1.120361, extrema: (-1.0, 4.0)
B:  mean: 0.911458  RMS: 1.058202, extrema: (-1.0, 3.0)

with the patch:

R:  mean: 0.010417  RMS: 0.595119, extrema: (-3.0, 3.0)
G:  mean: 0.000000  RMS: 0.595119, extrema: (-2.0, 3.0)
B:  mean: -0.08854  RMS: 0.535218, extrema: (-2.0, 2.0)


Contrary to my earlier assertion, I have discovered the area averaged 
images differ from the gimp on the bottom row of the image.

comparing area-averaged to gimp:

R:  mean: 0.041667  RMS: 0.204124, extrema: (0.0, 1.0)
G:  mean: 0.052083  RMS: 0.228218, extrema: (0.0, 1.0)
B:  mean: 0.020833  RMS: 0.144338, extrema: (0.0, 1.0)




douglas

-------------- next part --------------
A non-text attachment was scrubbed...
Name: pil-antialias.diff
Type: text/x-patch
Size: 2276 bytes
Desc: not available
Url : http://mail.python.org/pipermail/image-sig/attachments/20050513/38fefd50/pil-antialias.bin

From simon.clay at imperial.ac.uk  Fri May 13 19:33:45 2005
From: simon.clay at imperial.ac.uk (Simon Clay)
Date: Fri, 13 May 2005 18:33:45 +0100
Subject: [Image-SIG] Installing: "The _imaging C module is not installed"
Message-ID: <20050513173345.GA23794@imperial.ac.uk>


Hi.

  I'm trying to install PIL with python 2.3.3 on a Debian Linux box.
There are no errors from setup.py, and the summary looks like this:

--------------------------------------------------------------------
PIL 1.1.5 BUILD SUMMARY
--------------------------------------------------------------------
version       1.1.5
platform      linux2 2.3.3 (#1, May  6 2004, 13:40:17)
              [GCC 2.95.4 20011002 (Debian prerelease)]
--------------------------------------------------------------------
*** TKINTER support not available
--- JPEG support ok
--- ZLIB (PNG/ZIP) support ok
--- FREETYPE2 support ok
--------------------------------------------------------------------

Then I run selftest.py, and get:

*** The _imaging C module is not installed

Some quick googling turned up this list, and:
http://effbot.org/zone/pil-imaging-not-installed.htm
which talks about this very problem.  Following the instructions there,
all is well until:

>>> import _imaging
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
ImportError: /home/simon/python/tmp/Imaging-1.1.5/PIL/_imaging.so:
undefined symbol: __ctype_b_loc

I'm not quite sure where to go from here.  Any suggestions are
welcome...

	Simon.


From fredrik at pythonware.com  Fri May 13 21:06:21 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Fri, 13 May 2005 21:06:21 +0200
Subject: [Image-SIG] Installing: "The _imaging C module is not installed"
References: <20050513173345.GA23794@imperial.ac.uk>
Message-ID: <d62tcj$n9d$1@sea.gmane.org>

Simon Clay wrote:

> >>> import _imaging
> Traceback (most recent call last):
>   File "<stdin>", line 1, in ?
> ImportError: /home/simon/python/tmp/Imaging-1.1.5/PIL/_imaging.so:
> undefined symbol: __ctype_b_loc
>
> I'm not quite sure where to go from here.  Any suggestions are
> welcome...

judging from

    http://lists.debian.org/debian-glibc/2003/06/msg00062.html

your gcc install seems to be somewhat botched.

</F>




From eachand at gmail.com  Mon May  9 19:02:59 2005
From: eachand at gmail.com (Aman)
Date: Mon, 9 May 2005 10:02:59 -0700
Subject: [Image-SIG] Tif Handling routines
Message-ID: <5dc4aea70505091002aae2c46@mail.gmail.com>

Hello Folks
 Am new to python and trying to put up a prototype Tiff image splitter and 
merger could i get some one to help me start it off or if you have source 
codes it will be most welcomed.Pressed for time
 Cheers
Edmond
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.python.org/pipermail/image-sig/attachments/20050509/b2e6e908/attachment.html

From fredrik at pythonware.com  Fri May 13 22:06:43 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Fri, 13 May 2005 22:06:43 +0200
Subject: [Image-SIG] Tif Handling routines
References: <5dc4aea70505091002aae2c46@mail.gmail.com>
Message-ID: <d630tp$32n$1@sea.gmane.org>

"Aman" <eachand at gmail.com> wrote:

> Am new to python and trying to put up a prototype Tiff image splitter and
> merger could i get some one to help me start it off or if you have source
> codes it will be most welcomed.Pressed for time

if you're pressed for time, I suggest using the tiffsplit and tiffcp
utilities included in the libtiff package:

    http://www.libtiff.org/

you can use Python's subprocess module to run the commands
from Python.

</F>




From stuartaw at gmail.com  Mon May 16 03:37:08 2005
From: stuartaw at gmail.com (qzm)
Date: Mon, 16 May 2005 13:37:08 +1200
Subject: [Image-SIG] PIL 1.1.5+window 2000 problem.
Message-ID: <8046ffd605051518374b3cffbe@mail.gmail.com>

Just wondering if anyone has ever seen this.

I just hit a problem with the binary install of PIL1.1.5 on a windows
2000 machine with python 2.4.1,
when I load a tga it comes out as  an incorrect size, and nothing else works.

ie:
import Image
im=Image.open('test.tga')
print im.size

returns (2,0)
when it should be (640,480)

Works fine on most of the other machine around here :-/

Am going to try a rebuild from source next.
Does anyone know if PIL uses an inernal library for tga, and ifnot
what the external one is? perhaps that is damaged.

Never seen this one before :-/

Regards,
Stuart.

From fredrik at pythonware.com  Mon May 16 08:36:26 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Mon, 16 May 2005 08:36:26 +0200
Subject: [Image-SIG] PIL 1.1.5+window 2000 problem.
References: <8046ffd605051518374b3cffbe@mail.gmail.com>
Message-ID: <d69f0l$8o4$1@sea.gmane.org>

"qzm" <stuartaw at gmail.com> wrote:

> I just hit a problem with the binary install of PIL1.1.5 on a windows
> 2000 machine with python 2.4.1,
> when I load a tga it comes out as  an incorrect size, and nothing else works.
>
> ie:
> import Image
> im=Image.open('test.tga')
> print im.size
>
> returns (2,0)
> when it should be (640,480)
>
> Works fine on most of the other machine around here :-/

how did you transfer the file to the windows machine?  are you sure
didn't accidentally transfer it in text mode?

(if the same file works on other platforms, feel free to mail me a copy
so I can look at it).

> Am going to try a rebuild from source next.
> Does anyone know if PIL uses an inernal library for tga, and ifnot
> what the external one is? perhaps that is damaged.

PIL uses Python code to parse the TGA headers (TgaImagePlugin.py).

</F>




From ronaldoussoren at mac.com  Mon May 16 20:48:04 2005
From: ronaldoussoren at mac.com (Ronald Oussoren)
Date: Mon, 16 May 2005 20:48:04 +0200
Subject: [Image-SIG] setup.py and automaticly detecting freetype2
Message-ID: <393A6956-61DE-4740-9825-C1F95ADB8D4A@mac.com>

Hi,

The setup.py in Imageing-1.1.5 tries to autodetect if, and which,  
freetype is installed. This fails to consider the --include-dirs  
command-line option.

I've freetype2 (as well as libjpeg and libtiff) installed in /opt/ 
local, which is not on the default search-path of distutils. Hence I  
built the extensions using::

    python setup.py build_ext --include-dirs=/opt/local/include -- 
library-dirs=/opt/local/lib

This works for the detection of libjpeg and libtiff because that code  
looks at the library path of self.compiler. The freetype  
configuration code also tries to find the freetype2 include-file.  
This only looks at a path that's hardcoded in setup.py, and ignores  
the include-path of self.compiler. IMHO this code should
look at self.compiler.include_dirs instead of just include_dirs  
(setup.py around line 216)

Regards,
     Ronald
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2105 bytes
Desc: not available
Url : http://mail.python.org/pipermail/image-sig/attachments/20050516/e801ef94/smime.bin

From djarb at highenergymagic.org  Mon May 16 22:52:52 2005
From: djarb at highenergymagic.org (Daniel Arbuckle)
Date: Mon, 16 May 2005 13:52:52 -0700
Subject: [Image-SIG] PIL patch
Message-ID: <20050516205252.GA9761@highenergymagic.org>

The following patch should be applied to OleFileIO.py:

298c298
<             if ix == 0xFFFFFFFEL or x == 0xFFFFFFFFL:
---
>             if ix == 0xFFFFFFFEL or ix == 0xFFFFFFFFL:

From fredrik at pythonware.com  Tue May 17 08:09:42 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Tue, 17 May 2005 08:09:42 +0200
Subject: [Image-SIG] PIL patch
References: <20050516205252.GA9761@highenergymagic.org>
Message-ID: <d6c1q9$e9g$1@sea.gmane.org>

Daniel Arbuckle write;

> The following patch should be applied to OleFileIO.py:
>
> 298c298
> <             if ix == 0xFFFFFFFEL or x == 0xFFFFFFFFL:
> ---
> >             if ix == 0xFFFFFFFEL or ix == 0xFFFFFFFFL:

thanks, but you're almost two years late:

--- OleFileIO.py        (revision 260)
+++ OleFileIO.py        (revision 261)
@@ -11,6 +11,7 @@
 # History:
 # 1997-01-20 fl   Created
 # 1997-01-22 fl   Fixed 64-bit portability quirk
+# 2003-09-09 fl   Fixed typo in OleFileIO.loadfat (noted by Daniel Haertle)
 #
 # Notes:
 # FIXME: change filename to use "a/b/c" instead of ["a", "b", "c"]
@@ -295,7 +296,7 @@
         fat = []
         for i in range(0, len(sect), 4):
             ix = i32(sect, i)
-            if ix == 0xFFFFFFFEL or x == 0xFFFFFFFFL:
+            if ix == 0xFFFFFFFEL or ix == 0xFFFFFFFFL:
                 break
             s = self.getsect(ix)
             fat = fat + map(lambda i, s=s: i32(s, i), range(0, len(s), 4))

</F>




From franz at trispen.com  Fri May 20 13:33:07 2005
From: franz at trispen.com (Franz Struwig)
Date: Fri, 20 May 2005 13:33:07 +0200
Subject: [Image-SIG] PIL Plus and ImageCrackCode
Message-ID: <opsq2nphaibz1klc@snuifer.trispen.com>

Greetings,
I'm new to Python and trying to use PIL to:
a) Cross-correlate two images - i.e shift/scale/rotate the one image until  
they produce the best match.
b) Find the Center Of Gravity of each letter on a page.

Any advice or help in the above tasks would be greatly appreciated!

I see that the ImageCrackCode module is able to calculate the Center Of  
Gravity of "features" on a page.  Unfortunately I haven't been able to use  
the module - I don't think I have PIL Plus.

What is PIL Plus, and how can I get it?

Thanks,
Franz



From fredrik at pythonware.com  Fri May 20 16:12:23 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Fri, 20 May 2005 16:12:23 +0200
Subject: [Image-SIG] PIL Plus and ImageCrackCode
References: <opsq2nphaibz1klc@snuifer.trispen.com>
Message-ID: <d6kr69$8e5$1@sea.gmane.org>

Franz Struwig wrote:

> What is PIL Plus, and how can I get it?

PIL Plus (aka imToolkit) was a commercial extension to Python,
that was available to PIL Support customers.  the PIL Plus ex-
tension is no longer available:

    http://www.pythonware.com/products/pil/plus.htm

</F>




From andy_chamberlain at yahoo.co.uk  Tue May 24 17:41:45 2005
From: andy_chamberlain at yahoo.co.uk (Andrew Chamberlain)
Date: Tue, 24 May 2005 16:41:45 +0100 (BST)
Subject: [Image-SIG] PIL 1.1.6 (ImageMath)
Message-ID: <20050524154145.85414.qmail@web25208.mail.ukl.yahoo.com>

Hello,

I was just wondering when PIL 1.1.6 might be released (or even a beta
version of it)?  The reason I ask is because I'm trying to calculate a
ratio image between two bands in an RGB image, and was hoping the 
ImageMath module in 1.1.5 might be able to help.

I've looked at possibilities in 1.1.5, including the ImageChops module,
a division operator doesn't seem to be available at the moment.

Many thanks in advance,

Andy Chamberlain




		
___________________________________________________________ 
How much free photo storage do you get? Store your holiday 
snaps for FREE with Yahoo! Photos http://uk.photos.yahoo.com

From rzantow at ntelos.net  Wed May 25 03:16:16 2005
From: rzantow at ntelos.net (rzed)
Date: Tue, 24 May 2005 21:16:16 -0400
Subject: [Image-SIG] Regarding Wide lines in PIL
In-Reply-To: <20050524154145.85414.qmail@web25208.mail.ukl.yahoo.com>
References: <20050524154145.85414.qmail@web25208.mail.ukl.yahoo.com>
Message-ID: <4293D1E0.1070608@ntelos.net>

If I have the correct version of the PIL code, specifying the new width 
parameter to ImageDraw's line() function results in a line of twice the 
specified thickness. It appears to be because, in ImagingDrawWideLine 
(Draw.c), the width is both added to and subtracted from the line 
positions. If dx and dy were halved, the  specification should work 
correctly (within rounding limits, anyway).

     dx = (int) (d * (y1-y0) + 0.5);
     dy = (int) (d * (x1-x0) + 0.5);

     add_edge(e+0, x0 - dx,  y0 + dy, x1 - dx,  y1 + dy);
     add_edge(e+1, x1 - dx,  y1 + dy, x1 + dx,  y1 - dy);
     add_edge(e+2, x1 + dx,  y1 - dy, x0 + dx,  y0 - dy);
     add_edge(e+3, x0 + dx,  y0 - dy, x0 - dx,  y0 + dy);

To an extent, it can be worked around by halving the width 
specification, of course.

-- 
rzed

From douglas at paradise.net.nz  Wed May 25 04:36:49 2005
From: douglas at paradise.net.nz (Douglas Bagnall)
Date: Wed, 25 May 2005 14:36:49 +1200
Subject: [Image-SIG] PIL 1.1.6 (ImageMath)
In-Reply-To: <20050524154145.85414.qmail@web25208.mail.ukl.yahoo.com>
References: <20050524154145.85414.qmail@web25208.mail.ukl.yahoo.com>
Message-ID: <4293E4C1.7090407@paradise.net.nz>

Andrew Chamberlain wrote:
>  The reason I ask is because I'm trying to calculate a
> ratio image between two bands in an RGB image, and was hoping the 
> ImageMath module in 1.1.5 might be able to help.

If speed does not matter much, something like this should work:

im = Image.open('some rgb image')
im2 = Image.new('L', im.size)
SCALE = 128

ratios = [min(int(SCALE * r / (g + 0.01)), 255)
           for r, g, b in im.getdata()]

im2.putdata(ratios)
im2.save('/tmp/ratio.png')



douglas


From fredrik at pythonware.com  Wed May 25 08:50:57 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Wed, 25 May 2005 08:50:57 +0200
Subject: [Image-SIG] PIL 1.1.6 (ImageMath)
References: <20050524154145.85414.qmail@web25208.mail.ukl.yahoo.com>
Message-ID: <d71755$dt9$1@sea.gmane.org>

Andrew Chamberlain wrote:

> I was just wondering when PIL 1.1.6 might be released (or even a beta
> version of it)?  The reason I ask is because I'm trying to calculate a
> ratio image between two bands in an RGB image, and was hoping the
> ImageMath module in 1.1.5 might be able to help.

you can get a sneak alpha release here (source only):

    http://effbot.org/downloads/#Imaging

</F> 




From fredrik at pythonware.com  Wed May 25 22:42:56 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Wed, 25 May 2005 22:42:56 +0200
Subject: [Image-SIG] Regarding Wide lines in PIL
References: <20050524154145.85414.qmail@web25208.mail.ukl.yahoo.com>
	<4293D1E0.1070608@ntelos.net>
Message-ID: <d72nt1$ida$1@sea.gmane.org>

"rzed" <rzantow at ntelos.net> wrote:

> If I have the correct version of the PIL code, specifying the new width
> parameter to ImageDraw's line() function results in a line of twice the
> specified thickness. It appears to be because, in ImagingDrawWideLine
> (Draw.c), the width is both added to and subtracted from the line
> positions. If dx and dy were halved, the  specification should work
> correctly (within rounding limits, anyway).
>
>      dx = (int) (d * (y1-y0) + 0.5);
>      dy = (int) (d * (x1-x0) + 0.5);
>
>      add_edge(e+0, x0 - dx,  y0 + dy, x1 - dx,  y1 + dy);
>      add_edge(e+1, x1 - dx,  y1 + dy, x1 + dx,  y1 - dy);
>      add_edge(e+2, x1 + dx,  y1 - dy, x0 + dx,  y0 - dy);
>      add_edge(e+3, x0 + dx,  y0 - dy, x0 - dx,  y0 + dy);
>
> To an extent, it can be worked around by halving the width
> specification, of course.

I've attached the 'official' patch.

</F>

Index: libImaging/Draw.c
===================================================================
--- libImaging/Draw.c   (revision 2434)
+++ libImaging/Draw.c   (working copy)
@@ -674,7 +675,7 @@
         return 0;
     }

-    d = width / sqrt(dx*dx + dy*dy);
+    d = width / sqrt(dx*dx + dy*dy) / 2.0;

     dx = (int) (d * (y1-y0) + 0.5);
     dy = (int) (d * (x1-x0) + 0.5);




From TeXSatz at uni-bremen.de  Wed May 25 23:15:12 2005
From: TeXSatz at uni-bremen.de (LML)
Date: Wed, 25 May 2005 23:15:12 +0200
Subject: [Image-SIG] PIL 1.1.5 polygon filling off by one vertical pixel,
 alpha blended drawing?
Message-ID: <4294EAE0.8070701@uni-bremen.de>

Hi all,
the ImageDraw.Draw.polygon function does not draw as I expect it.
Deep down in the C-code some Floor, Ceil and +0.5 seem to combine
for a strange effect.

1. Any suggestions to actually draw within the outline only?
2. Why can't I draw with alpha-blending (it seems like I
    have to im.paste to achieve this -- so inefficient...)
3. Do I have to destruct the Draw instance with del explicitly
    (like it's done in the example code)
    or is the implicit destruction enough to get it freed?
4. Is ImagePath.Path more efficient or just syntactic sugar?

Code to show the effect:

#piltest.py
import Image, ImageDraw

def save(image, fname='save.png'):
     W = file(fname, 'wb')
     image.save(W)
     W.close()

def testpic():
     """Test polygon, line etc. drawing."""
     im = Image.new('RGBA', (100,100))
     D = ImageDraw.Draw(im)
     fi=(0xbf, 0xbf, 0xbf, 0x7f)
     D.line([(39,0),(39,100)],fill=fi) #vertical lines
     D.line([(50,0),(50,100)],fill=fi)
     D.line([(61,0),(61,100)],fill=fi)
     D.line([(0,39),(100,39)],fill=fi) #horizontal lines
     D.line([(0,50),(100,50)],fill=fi)
     D.line([(0,61),(100,61)],fill=fi)
     fi=(0xbf, 0x00, 0x00, 0x7f) #red
     ou=(0x7f, 0x00, 0x00, 0x7f)
     D.polygon([(40, 40), (40, 60), (60, 60), (60, 40)], fill=fi, 
outline=ou)
     D.polygon([(10, 10), (10, 30), (30, 30), (30, 10)], fill=fi, 
outline=ou)
     D.polygon([(70, 10), (71, 30), (91, 30), (90, 10)], fill=fi, 
outline=ou)
     fi=(0x00, 0xbf, 0x00, 0x7f) #green
     ou=(0x00, 0x7f, 0x00, 0x7f)
     D.polygon([(20, 70), (10, 80), (20, 90), (30, 80)], fill=fi, 
outline=ou)
     D.polygon([(81, 70), (70, 80), (80, 90), (91, 80)], fill=fi, 
outline=ou)
     D.polygon([(50, 40), (40, 50), (50, 60), (60, 50)], fill=fi, 
outline=ou)
     D.point([(0, 0)],fill='red') #mark origin
     save(im)

the resulting save.png is attached -- hopefully small enough ;-)

I intend to use the polygon function to draw land/coastlines
combined with satellite (point) data -- that's where the alpha
blending is supposed to enter the picture...

Thanks for your help.
\bye Lothar (Meyer-Lerbs, Bremen, Germany)



-------------- next part --------------
A non-text attachment was scrubbed...
Name: save.png
Type: image/png
Size: 1780 bytes
Desc: not available
Url : http://mail.python.org/pipermail/image-sig/attachments/20050525/ec945c03/save.png

From fredrik at pythonware.com  Thu May 26 08:54:08 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Thu, 26 May 2005 08:54:08 +0200
Subject: [Image-SIG] PIL 1.1.5 polygon filling off by one vertical pixel,
	alpha blended drawing?
References: <4294EAE0.8070701@uni-bremen.de>
Message-ID: <d73rms$2sn$1@sea.gmane.org>

"LML" wrote:

> the ImageDraw.Draw.polygon function does not draw as I expect it.
> Deep down in the C-code some Floor, Ceil and +0.5 seem to combine
> for a strange effect.

can you elaborate: from a quick look, the image seems to match
the code pretty well.

> 1. Any suggestions to actually draw within the outline only?
> 2. Why can't I draw with alpha-blending (it seems like I
>     have to im.paste to achieve this -- so inefficient...)

recent versions support RGBA-on-RGB drawing; from the CHANGES
document:

    + Added experimental RGBA-on-RGB drawing support.  To use RGBA
      colours on an RGB image, pass "RGBA" as the second string to
      the ImageDraw.Draw constructor.

> 3. Do I have to destruct the Draw instance with del explicitly
>     (like it's done in the example code)
>     or is the implicit destruction enough to get it freed?

no.  normal garbage collection works here.

> 4. Is ImagePath.Path more efficient or just syntactic sugar?

Path is more efficient if you need to scale/transform the data before
you draw.  If you're doing the calculations elsewhere, you can use
any input format.

</F>




From fredrik at pythonware.com  Thu May 26 10:13:15 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Thu, 26 May 2005 10:13:15 +0200
Subject: [Image-SIG] PIL 1.1.5 polygon filling off by one vertical pixel,
	alpha blended drawing?
References: <4294EAE0.8070701@uni-bremen.de> <d73rms$2sn$1@sea.gmane.org>
Message-ID: <d740b6$f4s$1@sea.gmane.org>


>> the ImageDraw.Draw.polygon function does not draw as I expect it.
>> Deep down in the C-code some Floor, Ceil and +0.5 seem to combine
>> for a strange effect.
>
> can you elaborate: from a quick look, the image seems to match
> the code pretty well.

oh, never mind.  I'll look into this.

</F> 




From Ryan.Hruska at inl.gov  Thu May 26 19:08:09 2005
From: Ryan.Hruska at inl.gov (Ryan C Hruska)
Date: Thu, 26 May 2005 11:08:09 -0600
Subject: [Image-SIG] PIL / gdal_translate
Message-ID: <OF65DD9C0F.C5EF07EC-ON8725700D.005C44D1-8725700D.005E21DD@inel.gov>

I am using PIL to process aerial  imagery and would like to translate it 
from the sensor's frame coordinate system to a world coordinate system 
(UTM) using GPS/INS data collected at the time exposure. I am using python 
to calculate the UTM coordinates for the image corners and perspective 
center based on the camera model, but I haven't had any luck applying this 
to the imagery. It was suggested that I use gdal_translate utility to 
accomplish this. 

Anyway, is it possible to call the gdal_translate.exe utility from python 
and pass it the required parameters? I am using the Pythonwin development 
environment. Any help or guidance would be appreciated.

Ryan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.python.org/pipermail/image-sig/attachments/20050526/3618345e/attachment.htm
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/gif
Size: 19670 bytes
Desc: not available
Url : http://mail.python.org/pipermail/image-sig/attachments/20050526/3618345e/attachment.gif

From fredrik at pythonware.com  Thu May 26 19:40:24 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Thu, 26 May 2005 19:40:24 +0200
Subject: [Image-SIG] PIL / gdal_translate
References: <OF65DD9C0F.C5EF07EC-ON8725700D.005C44D1-8725700D.005E21DD@inel.gov>
Message-ID: <d751ig$u7q$1@sea.gmane.org>

Ryan C Hruska wrote:
> I am using PIL to process aerial  imagery and would like to translate it
> from the sensor's frame coordinate system to a world coordinate system
> (UTM) using GPS/INS data collected at the time exposure. I am using python
> to calculate the UTM coordinates for the image corners and perspective
> center based on the camera model, but I haven't had any luck applying this
> to the imagery. It was suggested that I use gdal_translate utility to
> accomplish this.
>
> Anyway, is it possible to call the gdal_translate.exe utility from python
> and pass it the required parameters? I am using the Pythonwin development
> environment. Any help or guidance would be appreciated.

start here:

    http://www.python.org/dev/doc/devel/lib/module-subprocess.html

if you're using Python 2.2 or 2.3, you can download a subprocess
installer here:

    http://effbot.org/downloads/#subprocess

</F>




From john.ertl at fnmoc.navy.mil  Fri May 27 22:48:52 2005
From: john.ertl at fnmoc.navy.mil (Ertl, John)
Date: Fri, 27 May 2005 13:48:52 -0700
Subject: [Image-SIG] Error when trying to open a .ps file
Message-ID: <E338ADD616B66043824B9ABF5CA6EF2332C6BE@lanexc107p.fnmoc.navy.mil>

All,

I am trying to open a .ps file and then eventually convert it to a .png but
I cannot even open it.   If I use a postscript viewer I can view the image
so I think the image is O.K.  I am using Python 2.4 on RH ES 3.0 and PIL
1.1.5.  I have been doing some work with other formats it has been great so
I think PIL is installed correctly but I have just had no luck with .ps
files.

Any ideas?

>>> im = Image.open("test.ps")

Traceback (most recent call last):
  File "<pyshell#2>", line 1, in -toplevel-
    im = Image.open("test.ps")
  File
"/opt/global/python/python-2.4/lib/python2.4/site-packages/PIL/Image.py",
line 1730, in open
    return factory(fp, filename)
  File
"/opt/global/python/python-2.4/lib/python2.4/site-packages/PIL/ImageFile.py"
, line 82, in __init__
    self._open()
  File
"/opt/global/python/python-2.4/lib/python2.4/site-packages/PIL/EpsImagePlugi
n.py", line 277, in _open
    raise IOError, "cannot determine EPS bounding box"
IOError: cannot determine EPS bounding box

Thanks,

John C. Ertl
Fleet Numerical Meteorology & Oceanography Center
7 Grace Hopper Ave
Monterey, CA 93943
phone: (831) 656-5704
fax:   (831) 656-4363


From fredrik at pythonware.com  Sat May 28 13:23:27 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Sat, 28 May 2005 13:23:27 +0200
Subject: [Image-SIG] Error when trying to open a .ps file
References: <E338ADD616B66043824B9ABF5CA6EF2332C6BE@lanexc107p.fnmoc.navy.mil>
Message-ID: <d79k76$m3q$1@sea.gmane.org>

"Ertl, John" wrote:

> I am trying to open a .ps file and then eventually convert it to a .png but
> I cannot even open it.   If I use a postscript viewer I can view the image
> so I think the image is O.K.  I am using Python 2.4 on RH ES 3.0 and PIL
> 1.1.5.  I have been doing some work with other formats it has been great so
> I think PIL is installed correctly but I have just had no luck with .ps
> files.
>
> Any ideas?
>
> >>> im = Image.open("test.ps")
>
> Traceback (most recent call last):
>   File "<pyshell#2>", line 1, in -toplevel-
>     im = Image.open("test.ps")
>   File
> "/opt/global/python/python-2.4/lib/python2.4/site-packages/PIL/Image.py",
> line 1730, in open
>     return factory(fp, filename)
>   File
> "/opt/global/python/python-2.4/lib/python2.4/site-packages/PIL/ImageFile.py"
> , line 82, in __init__
>     self._open()
>   File
> "/opt/global/python/python-2.4/lib/python2.4/site-packages/PIL/EpsImagePlugi
> n.py", line 277, in _open
>     raise IOError, "cannot determine EPS bounding box"
> IOError: cannot determine EPS bounding box

PIL can only open EPS files (a variant of PS designed for embedding),
and it looks like your PS file isn't an EPS.

I suggest running it directly through Ghostscript (in batch mode), and
processing the output instead.

</F>




From 550283447739-0001 at t-online.de  Tue May 31 11:03:12 2005
From: 550283447739-0001 at t-online.de (Oliver Albrecht)
Date: Tue, 31 May 2005 11:03:12 +0200
Subject: [Image-SIG] Help colors
Message-ID: <429C2850.6010804@t-online.de>


Greetings,
I'm new to Python and trying to use PIL to:

- Wich Colorreduce Method used PIL other as "Web"?
	"Octree" , "Median" ?

- Why has the GIF format double size?  (I use PSP as Graphic-program )
- How discover the nearest color?
	x= (r1,g1,b1) - (r,g,b)
	y= (r2,g2,b2) - (r,g,b)
	if pos(x) > pos(y): nearest=y
	else: nearest=x

I want make a script to switch the colorpalette from images.
	# numerate all pixel(palette_Id)
	MapL=list(enumerate(data_list))
	# turn round nr/palette_Id
	MapL= [(b, [a]) for a, b in MapL]
	MapL.sort()
	Listshort=[("",[""])]
	# delete all double palette_Id but put the key(nr) to it
	for value, key in MapL:
     		if value != Listshort[-1][0]:
         		Listshort.append((value,key))
     		else:
         		Listshort[-1][1].extend(key)
	Listshort=sorted(Listshort[1:])

	# now load a other palette and compare this
	sorted(palette1,palette2)

	# then switch the palette_Ids in Listshort and reinsert it as im.data

If it give another, better way tell it me. else i will post the
completed script.

tiny-fault-report:
I think in the PIL-Beta-Docu this Example is an fault.
http://effbot.org/imagingbook/imagepalette.htm

##Getting the Palette Contents Using Resize/Convert
assert im.mode == "P"

lut = im.resize((256, 1))
lut.putdata(range(256))
lut = im.convert("RGB").getdata()
       ^^
# lut now contains a sequence of (r, g, b) tuples

"lut" has the sequence from "im" but not from "lut.putdata(range(256))"
"lut" is a copy from "im" but "im" is not resized.
-- 
<!--Olliminatore-->

From 550283447739-0001 at t-online.de  Tue May 31 15:00:52 2005
From: 550283447739-0001 at t-online.de (Oliver Albrecht)
Date: Tue, 31 May 2005 15:00:52 +0200
Subject: [Image-SIG] Help GIF format
Message-ID: <429C6004.4040602@t-online.de>

by Fredrik Lundh Jul 11 1998
> When writing GIF files, PIL uses a simpleminded encoding that, by some
> odd reason, LZW decoders have no trouble reading.  To write compressed
> GIF files, there are a number of options:
> 
>     -- install NETPBM, and hack GifImagePlugin so it uses _save_ppm instead
>        of _save (just remove "_ppm" from the latter).
>     -- write an _imaging_gif module that takes a PIL image and writes
>        a GIF version of it.  how you implement that is up to you...

_imaging_gif  has anyone this module?
I try compile my model with py2exe, but it hold on GifImagePlugin on 
line  ['_imaging_gif'] sould i delete this line!?

and a additional little color question:
how can i simple replace one color?
here my means:
         im_pal=[]
         img_pall=im.getpalette()
         for RGB in xrange(0,len(img_pall),3):
             pal_tmp=img_pall[RGB:RGB+3]
             if pal_tmp==[0, 0, 132]:
                 pal_tmp=[255, 0, 255]
             im_pal[RGB:RGB]=pal_tmp
         im.putpalette(im_pal)
-- 
<!--Olliminatore-->

From fredrik at pythonware.com  Tue May 31 15:43:00 2005
From: fredrik at pythonware.com (Fredrik Lundh)
Date: Tue, 31 May 2005 15:43:00 +0200
Subject: [Image-SIG] Help GIF format
References: <429C6004.4040602@t-online.de>
Message-ID: <d7hpft$6n7$1@sea.gmane.org>

Oliver Albrecht wrote:

> I try compile my model with py2exe, but it hold on GifImagePlugin on
> line  ['_imaging_gif'] sould i delete this line!?

I'm pretty sure you can tell py2exe to ignore optional extensions.  if you
get stuck, you can replace the try/import/except statement in GifImagePlugin
with a simple assignment:

    _imaging_gif = None

</F> 




From christianj at implicitnetworks.com  Wed May 25 07:24:02 2005
From: christianj at implicitnetworks.com (Christian M. Jensen)
Date: Wed, 25 May 2005 05:24:02 -0000
Subject: [Image-SIG] PsdImagePlugin.py
Message-ID: <326A1C8DE34E8E4189BE602CE791678403F3B4@corporate.implicit.implicitnetworks.com>

Anyone have any issues reading PSDs from CS?

 

im = Image.open("test2.psd")

  File "C:\Python24\Lib\site-packages\PIL\Image.py", line 1745, in open

    raise IOError("cannot identify image file")

IOError: cannot identify image file

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.python.org/pipermail/image-sig/attachments/20050525/d4aee152/attachment.htm