open .raw image data using python

Question

I have been searching google for the method to display a raw image data using python libraries but couldn't find any proper solution. The data is taken from a camera module and it has the '.raw' extension. Also when I tried to open it in the terminal via 'more filename.raw', the console said that this is a binary file. Vendor told me that the camera outputs 16-bits raw greyscale data.

But I wonder how I can display this data via PIL, Pillow or just Numpy. I have tested the PIL's Image module. However, it couldn't identify the image data file. It seems the PIL doesn't consider the .raw file as an image data format. BMP files could be displayed, but this '.raw' couldn't.

Also when I tried with just read function and matplotlib, like the followings

from matplotlib import pyplot as plt
f = open("filename.raw", "rb").read() 
plt.imshow(f) 
plt.show()

then an error occurs with

ERROR: Image data can not convert to float

Any idea will be appreciated.

link: camera module

I made some improvement with the following codes. But now the issue is that this code displays only some portion of the entire image.

from matplotlib import pyplot as plt
import numpy as np
from StringIO import StringIO
from PIL import *
scene_infile = open('G0_E3.raw','rb')
scene_image_array = np.fromfile(scene_infile,dtype=np.uint8,count=1280*720)
scene_image = Image.frombuffer("I",[1280,720],
                                 scene_image_array.astype('I'),
                                 'raw','I',0,1)
plt.imshow(scene_image)
plt.show()

RAW is no image format in the sense a bitmap or jpeg is. These images contain color per pixel for example in rgb values. RAW basically includes the raw data of the photo ship, that has not been converted to color values yet (hence the 16-bit greyscale). This is why it is often used by (semi-)professional photographers to be able to alter the way colors are displayed in post-processing. Since the step of converting the data to color per pixel has not been done yet, all standard picture libraries will most likely not be able to read the RAW data format. — Dux
– Dux, Commented Sep 7, 2015 at 14:05
Have you searched SO for numpy and RAW? I recall something about RAW having a brand specific header block. — hpaulj
– hpaulj, Commented Sep 7, 2015 at 14:29
you might want to look at rawkit. I've never used it so can't comment on its usefulness to you, but it looks like it might be helpful here — tmdavison
– tmdavison, Commented Sep 7, 2015 at 14:29
There also exists a program (and library) which can read and convert RAW images from multiple camera models, see cybercom.net/~dcoffin/dcraw — Michiel Overtoom
– Michiel Overtoom, Commented Sep 7, 2015 at 14:33
Thank you for all the answers. But I wanted to understand how to open and manipulate the raw image manually by python. And I should have removed the 'from StringIO...' line. I don't use it now. — Nownuri
– Nownuri, Commented Sep 8, 2015 at 8:47

jds · Accepted Answer · 2017-08-13 18:19:36Z

24

Have a look at rawpy:

import rawpy
import imageio

path = 'image.raw'
raw = rawpy.imread(path)
rgb = raw.postprocess()
imageio.imsave('default.tiff', rgb)

rgb is just an RGB numpy array, so you can use any library (not just imageio) to save it to disk.

If you want to access the unprocessed Bayer data, then do:

bayer = raw.raw_image

Comments

Nice · Accepted Answer · 2021-07-01 02:36:05Z

import numpy as np
from PIL import Image
import rawpy


# For Adobe DNG image
input_file = 'path/to/rawimage.dng'
rawimg = rawpy.imread(input_file)
npimg = rawimg.raw_image


# For raw image (without any headers)
input_file = 'path/to/rawimage.raw'
npimg = np.fromfile(input_file, dtype=np.uint16)
imageSize = (3648, 2736)
npimg = npimg.reshape(imageSize)


# Save the image from array to file.
# TIFF is more suitable for 4 channel 10bit image
# comparing to JPEG
output_file = 'out.tiff'
Image.fromarray(npimg/1023.0).save(output_file)

Collectives™ on Stack Overflow

open .raw image data using python

2 Answers 2

Comments

Comments

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Linked

Related