PIL FITs image decoder

[jbrewer]

I'm trying to read in a FITs image file for my research, and I decided
that writing a file decoder for the Python imaging library would be the
easiest way to accomplish this for my needs. FITs is a raw data format
used in astronomy.

Anyway, I followed the example in the PIL documentation online, and I
also had a look at the FITs image stub file included with PIL 1.1.5. I
cooked up something that should work (or nearly work), but I keep
running into either one of two errors (below). The crux of the problem
appears to be that my plugin isn't being registered properly and isn't
being read at runtime when I call Image.open().

1.) The library loads the FitsStubImagePlugin.py file from the
site-packages directory (instead of my plugin) and then gives the
following error:

File
"/System/Library/Frameworks/Python.framework/Versions/2.3/lib/python2.3/site-packages/PIL/ImageFile.py",
line 255, in load
raise IOError("cannot find loader for this %s file" % self.format)
IOError: cannot find loader for this FITS file

2.) I remove the FitsStubImagePlugin.py, pyc files and stick my plugin
in the directory instead (my plugin was already in the PYTHONPATH
before). Then I get the following error:

Traceback (most recent call last):
File "FitsImagePlugin.py", line 111, in ?
image = Image.open(fits_name).save(jpg_name)
File
"/System/Library/Frameworks/Python.framework/Versions/2.3/lib/python2.3/site-packages/PIL/Image.py",
line 1745, in open
raise IOError("cannot identify image file")
IOError: cannot identify image file

This seems like I'm either doing (or not doing) something really stupid
or the docs are outdated. Any help would be greatly appreciated!

Jeremy

------------------------------ Code below
-----------------------------------

import Image, ImageFile

_handler = None

##
# Install application-specific FITS image handler.
#
# @param handler Handler object.

def register_handler(handler):
global _handler
_handler = handler

# --------------------------------------------------------------------
# Image adapter

def _accept(prefix):
return prefix[:6] == "SIMPLE"

class FitsImageFile(ImageFile.StubImageFile):
#class FitsImageFile(ImageFile.ImageFile):

format = "FITS"
format_description = "FITs raw image"

def _open(self):

# byte offset for FITs images
byte_offset = 2880

# check header for valid FITs image
header_data = self.fp.read(byte_offset)

# headers are stored in 80 character strings, so sort them out
into a
# nice list
i = 0
headers = []
while header_data[i] != "\n" and i < len(header_data):
headers.append(header_data[i:i + 80])
i += 81

# parse individual headers
ok = False
for header in headers:
words = header.split()

try:
keyword = words[0]
value = words[2]
except IndexError:
ok = False
break

if keyword == "NAXIS" and value == 2:
ok = True
elif keyword == "NAXIS1":
xsize = value
elif keyword == "NAXIS2":
ysize = value

if not ok:
raise ValueError("file is not a valid FITs image")

# size in pixels (width, height)
self.size = (xsize, ysize)

# mode setting is always greyscale
self.mode = "F"

# data descriptor
self.tile = [("raw", (0, 0) + self.size, byte_offset, ("F", 0,
-1))]

# is this needed?
loader = self._load()
if loader:
loader.open(self)

def _load(self):
return _handler

def _save(im, fp, filename):
if _handler is None or not hasattr("_handler", "save"):
raise IOError("FITS save handler not installed")
_handler.save(im, fp, filename)

# register the plugin
Image.register_open(FitsImageFile.format, FitsImageFile, _accept)
Image.register_save(FitsImageFile.format, _save)

# method given in docs
#Image.register_open(FitsImageFile.format, FitsImageFile)

Image.register_extension(FitsImageFile.format, ".fits")
Image.register_extension(FitsImageFile.format, ".fit")

# test code
if __name__ == "__main__":
import os

fits_name = "fpC-001739-r1-0304.fit"
jpg_name = os.path.splitext(fits_name)[0] + ".jpg"

image = Image.open(fits_name).save(jpg_name)
print "Converted FITs image '%s' to jpeg image '%s'" % (fits_name,
jpg_name)

[Fredrik Lundh]

"jbrewer" wrote:

I'm trying to read in a FITs image file for my research, and I decided
that writing a file decoder for the Python imaging library would be the
easiest way to accomplish this for my needs. FITs is a raw data format
used in astronomy.

Anyway, I followed the example in the PIL documentation online, and I
also had a look at the FITs image stub file included with PIL 1.1.5. I
cooked up something that should work (or nearly work), but I keep
running into either one of two errors (below). The crux of the problem
appears to be that my plugin isn't being registered properly and isn't
being read at runtime when I call Image.open().

1.) The library loads the FitsStubImagePlugin.py file from the
site-packages directory (instead of my plugin) and then gives the
following error:

unfortunately, the stub loaders (which were introduced in 1.1.5) may over-
ride more specific plugins. this will be fixed in 1.1.6.

I don't see any obvious problems with your file, except that you should in-
herit from ImageFile, not StubImageFile:

class FitsImageFile(ImageFile.StubImageFile):

should be

class FitsImageFile(ImageFile.ImageFile):

but I assume you've already tried that.

if you can you mail me (or point me to) a sample file, I can take a look.

</F>

[Robert Kern]

jbrewer wrote:

I'm trying to read in a FITs image file for my research, and I decided
that writing a file decoder for the Python imaging library would be the
easiest way to accomplish this for my needs. FITs is a raw data format
used in astronomy.

http://www.stsci.edu/resources/software_hardware/pyfits

--
Robert Kern
ro*********@gmail.com

"In the fields of hell where the grass grows high
Are the graves of dreams allowed to die."
-- Richard Harter

[jbrewer]

>http://www.stsci.edu/resources/software_hardware/pyfits

I know and love PyFits, but I need to be able to do draw shapes on a
FITs image (and perhaps some other operations), and I don't believe
that PyFits allows these kinds of operations. It might be possible to
import the data into a numarray object and turn that into something PIL
can read, though.

Jeremy

[Robert Kern]

jbrewer wrote:

http://www.stsci.edu/resources/software_hardware/pyfits

I know and love PyFits, but I need to be able to do draw shapes on a
FITs image (and perhaps some other operations), and I don't believe
that PyFits allows these kinds of operations. It might be possible to
import the data into a numarray object and turn that into something PIL
can read, though.

If you can bear having two copies in memory, Image.frombuffer()
generally does the trick.

--
Robert Kern
ro*********@gmail.com

"In the fields of hell where the grass grows high
Are the graves of dreams allowed to die."
-- Richard Harter

[jbrewer]

>If you can bear having two copies in memory, Image.frombuffer()

generally does the trick.

What arguments do you pass to this function, and do you flatten the
array from the FITs image? I this but got garbage out for the image.

Jeremy

[jbrewer]

>If you can bear having two copies in memory, Image.frombuffer()

generally does the trick.

Also, does PIL have a contrast / scale option that is similar to zscale
in ds9 or equalize in Image Magick?

Jeremy

[Robert Kern]

jbrewer wrote:

[I wrote:]

If you can bear having two copies in memory, Image.frombuffer()
generally does the trick.

What arguments do you pass to this function, and do you flatten the
array from the FITs image? I this but got garbage out for the image.

The array would have to be contiguous, which it may not be immediately
coming out of PyFITS. You also need to be sure that you are using the
right mode for the data.

In [7]: import Image

In [8]: import numarray

In [9]: z = numarray.zeros((256,256)).astype(numarray.UInt8)

In [10]: z[128,:] = 255

In [11]: img = Image.frombuffer('L', (256,256), z)

That creates a black image with a white, horizontal line in the middle.

--
Robert Kern
ro*********@gmail.com

"In the fields of hell where the grass grows high
Are the graves of dreams allowed to die."
-- Richard Harter

[jbrewer]

I tried following your simple example (I already had something similar)
but with no luck. I'm completely stumped as to why this doesn't work.
I even tried manually scaling the data to be in the range 0-255 out of
desperation. The data is definitely contiguous and 32 bit floating
point. At this point I've tried everything I can think of. Any ideas?

Jeremy

# open the fits file using pyfits
fits = pyfits.open(fitsfile)
xsize, ysize = fits[0].data.shape
data = fits[0].data.astype("Float32")
fits.close()

# now create an image object
image = Image.frombuffer("F", (xsize, ysize), data)

# scale the data be 256 bit unsigned integers
#int_data = numarray.zeros((xsize, ysize)).astype("UInt8")
#min = data.min()
#max = data.max()
#for i in arange(xsize):
# for j in arange(ysize):
# scaled_value = (data[i, j] - min) * (255.0 / (max - min))
# int_data[i, j] = int(scaled_value + 0.5)

#image = Image.frombuffer("L", (xsize, ysize), int_data)