Convert float NumPy array to big endian

Question

I wrote a script and in the end I need to convert this array which is in type float64 to big endian int (>2i):

[[ 0.92702157  1.03092008  0.9072934  ...,  0.71617331  1.02524888
    1.07284994]
[ 0.99573712  0.96416766  0.9230931  ...,  0.66935196  0.64930711
0.5357821 ]
[ 0.98846306  1.03608056  0.79976885 ...,  0.69383804  0.62434976
0.88219911]
..., 
[ 0.91196013  0.87880101  0.97145563 ...,  0.79110817  1.19651477
0.98244941]
[ 1.0129829   0.81045263  0.95434107 ...,  0.99752385  1.08271169
1.12872492]
[ 0.94037117  0.81365084  0.94384051 ...,  0.82754351  1.03742172  1.]]

How I can do that?

Here is whole script

import numpy as np
import pyfits
from matplotlib import pyplot as plt
import glob
import os
import re
from struct import unpack,pack

global numbers
numbers=re.compile(r'(\d+)')


def numericalSort(value):
    parts = numbers.split(value)
    parts[1::2] = map(int, parts[1::2])
    return parts

dark=sorted(glob.glob(' *.fits'),key=numericalSort)
flat=sorted(glob.glob('/ *.fits'),key=numericalSort)
img=sorted(glob.glob('/ *.fits'),key=numericalSort)

#dark
sumd0 = pyfits.open(dark[0])
sumdd=sumd0[0].data
sumdd = sumdd.astype(float,copy=False)

for i in range(1,len(dark)):
    sumdi=pyfits.open(dark[i])
    sumdi=sumdi[0].data
    sumdd=sumdd+sumdi.astype(float, copy=False)


dd=sumdd/len(dark)


#flat
sumf0 = pyfits.open(flat[0])
sumff=sumf0[0].data
sumff = sumff.astype(float, copy=False)

for i in range(1,len(flat)):
    sumfi=pyfits.open(flat[i])
    sumfi=sumfi[0].data
    sumff=sumff+sumfi.astype(float,copy=False)

ff=sumff/len(flat)


df=(ff-dd)
maxx=np.max(df)
df=np.clip(df,1,maxx)


for n in range(len(img)):
    im=pyfits.open(img[n])
    imgg=im[0].data
    header=im[0].header
    imgg=imgg.astype(float,copy=False)
    x,y=im[0].shape
    m=np.max(imgg)
    imgg=np.clip((imgg-dd),1,m)
    imgg=imgg/df
    imgg=np.clip(imgg,0.5,1.5)


    #print imgg.dtype
    #imgg=imgg[200:950,150:1250]
    #imgg=imgg[::-1,:y]
     hdu = pyfits.PrimaryHDU(imgg,header)
     hdulist = pyfits.HDUList([hdu])
     hdulist.writeto('/c'+img[n][48:])
     plt.imshow(imgg,cmap=plt.cm.Greys_r)
#plt.savefig('/+'.png')
plt.show()

On last 9 line I need convert array imgg from float64 to >i2

You could try `a.view('>2i')` (where `a` is the name of your array) - you'll get back a view of the array with twice as many elements and an extra dimension. Would this be what you want? — Alex Riley, Jul 21 '15 at 18:12
No...I need save this array to data file...it is image in type >2i...At the begin I have image data in fits file (2.8MiB)...My script make some operation (+ - /) with this data array but this operations want type float...But when I save this array after I get 11 MiB of size... I need convert this array back to >2i — Franta Konopnik, Jul 21 '15 at 18:21

score 2 · Answer 1 · answered Jul 21 '15 at 18:18

2

>>> import numpy as np
>>> big_end = bytes(chr(0) + chr(1) + chr(3) + chr(2), 'utf-8')
>>> np.ndarray(shape=(2,),dtype='>i2', buffer=big_end)
array([  1, 770], dtype=int16)

Convert float NumPy array to big endian

1 Answers1

Linked