[Image-SIG] going between numpy array and PIL.Image not behaving as expected

[Ned Batchelder]

Yes, it definitely looks like a bug in fromarray.  typestr is '<i4' in 
my environment, and the check against typestr[1:] seems to recognize 
this.  The line

    typestr = typestr[:2]

should change, perhaps to:

    typestr = typestr[-2:]

--Ned.
http://nedbatchelder.com

Jim Vickroy wrote:
> Ned Batchelder wrote:
>> I'm not sure exactly what you are trying to do here, but the issue 
>> has to do with the mapping of numpy array elements into pixels.  
> Thanks for your clear and detailed reply and sorry for the vagueness 
> of my posting; you did correctly read my mind!
>> Your code uses 32-bit ints, and fromarray defaults to "L" mode, which 
>> is 8-bit grayscale pixels.  fromarray uses the shape of the array to 
>> create the shape of the image, but then just reads bytes until the 
>> image has all the data it needs.  In your case, it only needs to read 
>> 3 32-bit ints to get enough bytes to fill the 3x4 "L" mode image.  In 
>> the first three ints, the min byte is zero and the max byte is 2, 
>> which your image extrema verifies.
>>
>> If you change your code to use this:
>>
>>     source  = numpy.arange(0,12,dtype=numpy.int8)
>>
>> then everything will match up: your array has byte elements, and your 
>> image will have byte pixels.
> Thanks, this does indeed work and your explanation made me wonder why 
> specifying dtype=int (as in my posted script) did not work. 
>
> Here is the signature of PIL.Image.fromarray in my installation (v 1.1.6):
>
>     * fromarray(obj, mode=None)
>
> so when mode is not specified, the procedure determines it from the 
> attributes of "obj" as follows:
>     if mode is None:
>         typestr = arr['typestr']
>         if not (typestr[0] == '|' or typestr[0] == _ENDIAN or
>                 typestr[1:] not in ['u1', 'b1', 'i4', 'f4']):
>             raise TypeError("cannot handle data-type")
>         typestr = typestr[:2] ##### why isn't this: typestr = 
> typestr[1:] or typestr = typestr[1:3] ? ##################
>         if typestr == 'i4':
>             mode = 'I'
>         elif typestr == 'f4':
>             mode = 'F'
>         elif typestr == 'b1':
>             mode = '1'
>         elif ndim == 2:
>             mode = 'L'
>         elif ndim == 3:
>             mode = 'RGB'
>         elif ndim == 4:
>             mode = 'RGBA'
>         else:
>             raise TypeError("Do not understand data.")
>
> I  am  relatively inexperienced with both PIL and numpy, but the  
> statement:
>
>     * typestr = typestr[:2]
>
> seems to be incorrect; after it is executed, I do not see how typestr 
> can ever be any of ('i4' , 'f4', 'b1').
>
> If I make the indicated change to
>
>     * typestr = typestr[1:]
>
> then "mode" is correctly inferred from "obj" and I do not have to 
> explicitly specify it when applying the fromarray() procedure.
>
> Is this a logic error in fromarray()?
>>
>
>>
>> --Ned.
>
>>
>> Jim Vickroy wrote:
>>> Hello all,
>>>
>>> I am having no success getting numpy and PIL to behave as expected 
>>> when starting with a numpy array (see the attached script).
>>>
>>> Here is the output on my computer:
>>>
>>> <output>
>>> Python version: 2.5.2 (r252:60911, Feb 21 2008, 13:11:45) [MSC 
>>> v.1310 32 bit (Intel)]
>>> numpy version:  1.2.1
>>> PIL version:    1.1.6
>>> numpy source array:
>>> [[ 0  1  2  3]
>>> [ 4  5  6  7]
>>> [ 8  9 10 11]]
>>> numpy source array shape: (3, 4)
>>> PIL image size: (4, 3)
>>> Traceback (most recent call last):
>>>  File "C:\Documents and Settings\jim.vickroy\My 
>>> Documents\Projects\GOES\SXI\__trials__\numpy-PIL.py", line 30, in 
>>> <module>
>>>    ''' % (extrema, image.getextrema())
>>> AssertionError:
>>>   numpy image extrema (minimum,maximum): (0, 11)
>>>   PIL   image extrema (minimum,maximum): (0, 2)
>>> </output>
>>>
>>>
>>> I would appreciate pointers on what I'm doing incorrectly.
>>>
>>> Thanks,
>>> -- jv
>>> ------------------------------------------------------------------------
>>>
>>> _______________________________________________
>>> Image-SIG maillist  -  Image-SIG at python.org
>>> http://mail.python.org/mailman/listinfo/image-sig
>>>   
>>
>> -- 
>> Ned Batchelder, http://nedbatchelder.com
>>   
>

-- 
Ned Batchelder, http://nedbatchelder.com

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/image-sig/attachments/20081121/8cb96ab1/attachment-0001.htm>