Opening raw images on Python resulting in a different image compared to ImageJ

Question

I wrote this script to open a raw image and do some processing.

import numpy as np 
import matplotlib.pyplot as plt
PATH = "C:\Users\Documents\script_testing_folder\\"
IMAGE_PATH = PATH +"simulation01\\15x15_output_det_00001_raw_df_00000.bin"
raw_image = np.fromfile(IMAGE_PATH, dtype=np.uint64)
raw_image.shape = (15,15)
plt.imshow(raw_image,cmap = 'gray')
total_intensity = ndimage.sum(raw_image)
print total_intensity
plt.show()

Using this script I get an Image such as this:

In contrast... when I open the same image on ImageJ(file>import>raw (64bit real, 15x15 length and width)) I have this:

I have tried looking around, but I am not sure where I am going wrong when trying to reproduce the same image on python. Any help would be greatly appreciated.

Additionally when I sum the intensity in the image using:

total_intensity = ndimage.sum(raw_image)
print total_intensity

y I get 4200794456581938015, whereas, on ImageJ I get 0.585.

I am not sure where I am going wrong on these steps...

Thanks !

Edit: The original file if anyone wants to reproduce the results I got https://www.dropbox.com/s/po82z4uf2ku7k0e/15x15_output_det_00001_raw_df_00000.bin?dl=0

What format has your image file? Could you upload it somewhere? — hitzg, Jul 29 '15 at 16:49
Not sure, but my initial guess is that you want to change the dtype to np.float64 since ImageJ uses what is called 64-bit real. Also, notice that your intensity sum in ImageJ is a decimal. — user1767344, Jul 29 '15 at 17:01
https://www.dropbox.com/s/po82z4uf2ku7k0e/15x15_output_det_00001_raw_df_00000.bin?dl=0 The file that I was trying to process — WingZ_of_Silence, Jul 29 '15 at 17:02
`raw_image = np.fromfile(IMAGE_PATH, dtype=np.uint64)`: does this load your data in the right way? It might not give you an error if it does not, as it just reads in the binary data and interprets it somehow. I have never seen `.bin` images before. There is a way to load 16bit tif if that helps: http://stackoverflow.com/a/18483020/2156909 — P.R., Jul 29 '15 at 17:30
also check out this post about loading raw image data: http://stackoverflow.com/a/10902279/2156909 — P.R., Jul 29 '15 at 17:33

score 4 · Accepted Answer · answered Jul 29 '15 at 18:01

The problem is the endianness of your data (the order of the single bytes of a 64bit float). Fortunately, numpy has the functionality to solve this issue:

import numpy as np 
import matplotlib.pyplot as plt

# load the image
raw_image = np.fromfile('15x15_output_det_00001_raw_df_00000.bin')
raw_image = np.reshape(raw_image, (15, 15))

# swap the byte order
raw_image = raw_image.byteswap()

# output the sum of the intensities to check
total_intensity = np.sum(raw_image)
print "total intensity:", total_intensity

# plot the image
plt.imshow(raw_image,cmap = 'gray', interpolation='nearest')
plt.show()

Output:

total intensity: 0.585123878711

Result:

Opening raw images on Python resulting in a different image compared to ImageJ

1 Answers1