I am converting pdf files to image using Wand. Then, I do further image processing using ndimage.
I would like to directly convert the Wand image into a ndarray... I have seen the answer here, but it use OpenCV. Is this possible without using OpenCV?
For the moment I save a temporary file, which is re-opened with scipy.misc.imread()
As of Wand 0.5.3 this is supported directly
import numpy as np
from wand.image import Image
with Image(filename='rose:') as img:
array = np.array(img)
print(array.shape) #=> (70, 46, 3)
You can use a buffer, like this:
import cStringIO
import skimage.io
from wand.image import Image
import numpy
#create the image, then place it in a buffer
with Image(width = 500, height = 100) as image:
image.format = 'bmp'
image.alpha_channel = False
img_buffer=numpy.asarray(bytearray(image.make_blob()), dtype=numpy.uint8)
#load the buffer into an array
img_stringIO = cStringIO.StringIO(img_buffer)
img = skimage.io.imread(img_stringIO)
img.shape
Here is a Python3 version of MrMartin's answer
from io import BytesIO
import skimage.io
from wand.image import Image
import numpy
with Image(width=100, height=100) as image:
image.format = 'bmp'
image.alpha_channel = False
img_buffer = numpy.asarray(bytearray(image.make_blob()), dtype='uint8')
bytesio = BytesIO(img_buffer)
img = skimage.io.imread(bytesio)
print(img.shape)
Here's what worked for me with Python3 without buffers:
from wand.image import Image
import numpy
with Image(width=100, height=100) as image:
image.format = 'gray' #If rgb image, change this to 'rgb' to get raw values
image.alpha_channel = False
img_array = numpy.asarray(bytearray(image.make_blob()), dtype='uint8').reshape(image.size)
