How to read image from in memory buffer (StringIO) or from url with opencv python library - image

Question

Just share a way to create opencv image object from in memory buffer or from url to improve performance.
Sometimes we get image binary from url, to avoid additional file IO, we want to imread this image from in memory buffer or from url, but imread only supports read image from file system with path.

Answer 1

To create an OpenCV image object with in memory buffer(StringIO), we can use OpenCV API imdecode, see code below:
import cv2
import numpy as np
from urllib2 import urlopen
from cStringIO import StringIO
def create_opencv_image_from_stringio(img_stream, cv2_img_flag=0):
img_stream.seek(0)
img_array = np.asarray(bytearray(img_stream.read()), dtype=np.uint8)
return cv2.imdecode(img_array, cv2_img_flag)
def create_opencv_image_from_url(url, cv2_img_flag=0):
request = urlopen(url)
img_array = np.asarray(bytearray(request.read()), dtype=np.uint8)
return cv2.imdecode(img_array, cv2_img_flag)

Answer 2

As pointed out in the comments to the accepted answer, it is outdated and no longer functional.
Luckily, I had to solve this very problem using Python 3.7 with OpenCV 4.0 recently.
To handle image loading from an URL or an in-memory buffer, I defined the following two functions:
import urllib.request
import cv2
import numpy as np
def get_opencv_img_from_buffer(buffer, flags):
bytes_as_np_array = np.frombuffer(buffer.read(), dtype=np.uint8)
return cv2.imdecode(bytes_as_np_array, flags)
def get_opencv_img_from_url(url, flags):
req = urllib.request.Request(url)
return get_opencv_img_from_buffer(urllib.request.urlopen(req), flags)
As you can see one depends on the other.
The first one, get_opencv_img_from_buffer, can be used to get an image object from an in-memory buffer. It assumes that the buffer has a read method and that it returns an instance of an object that implements the buffer protocol
The second one, get_opencv_img_from_url, generates an image directly from an URL.
The flags argument is passed on to cv2.imdecode,
which has the following constants predefined in cv2:
cv2.IMREAD_ANYCOLOR - If set, the image is read in any possible color format.
cv2.IMREAD_ANYDEPTH - If set, return 16-bit/32-bit image when the input has the corresponding depth, otherwise convert it to 8-bit.
cv2.IMREAD_COLOR - If set, always convert image to the 3 channel BGR color image.
cv2.IMREAD_GRAYSCALE - If set, always convert image to the single channel grayscale image (codec internal conversion).
cv2.IMREAD_IGNORE_ORIENTATION - If set, do not rotate the image according to EXIF's orientation flag.
cv2.IMREAD_LOAD_GDAL - If set, use the gdal driver for loading the image.
cv2.IMREAD_REDUCED_COLOR_2 - If set, always convert image to the 3 channel BGR color image and the image size reduced 1/2.
cv2.IMREAD_REDUCED_COLOR_4 - If set, always convert image to the 3 channel BGR color image and the image size reduced 1/4.
cv2.IMREAD_REDUCED_COLOR_8 - If set, always convert image to the 3 channel BGR color image and the image size reduced 1/8.
cv2.IMREAD_REDUCED_GRAYSCALE_2 - If set, always convert image to the single channel grayscale image and the image size reduced 1/2.
cv2.IMREAD_REDUCED_GRAYSCALE_4 - If set, always convert image to the single channel grayscale image and the image size reduced 1/4.
cv2.IMREAD_REDUCED_GRAYSCALE_8 - If set, always convert image to the single channel grayscale image and the image size reduced 1/8.
cv2.IMREAD_UNCHANGED - If set, return the loaded image as is (with alpha channel, otherwise it gets cropped).