How to convert image which type is bytes to numpy.ndarray?

Question

I'm trying to optimize my code.

First, I get an image, which type is bytes

Then I have to write that image to file system.

with open('test2.jpg', 'wb') as f:
    f.write(content)

Finally I read this image with

from scipy import misc
misc.imread('test2.jpg')

which convert image to np.array.

I want to skip part where I write image to file system, and get np.array.

P.S.: I tried to use np.frombuffer(). It doesn't work for me, cause two np.arrays are not the same. Convert str to numpy.ndarray

For test you can try yourself:

file = open('test1.jpg', 'rb')
content = file.read()

content is image which type is bytes. I get it from API request. — Samat Sadvakasov, Apr 02 '19 at 10:56
Something like that \xac\x03\x86_\x8a\x04>\x83\xfe\xa2\xf2T\x0e\xb59\x14p\xc5A\xe7<\xd4\x12\x08\xe0t\xa90J\x14\xfa\x89\xcf\xf4\x — Samat Sadvakasov, Apr 02 '19 at 10:58
which API? To get a meanigful answer include a piece of working code that others can try. — DrBwts, Apr 02 '19 at 10:58
the struct library might be waht you are looking for https://docs.python.org/2/library/struct.html — DrBwts, Apr 02 '19 at 11:01
Take a look at https://stackoverflow.com/questions/44324944/how-to-decode-jpg-image-from-memory — Warren Weckesser, Apr 02 '19 at 13:55

score 3 · Accepted Answer · edited Jun 20 '20 at 09:12

3

My first answer in rap...

Wrap that puppy in a BytesIO

And away you go

So, to generate some synthetic data similar to what you get from the API:

file = open('image.jpg','rb')
content = file.read()

That looks like this which has all the hallmarks of a JPEG:

content = b'\xff\xd8\xff\xe0\x00\x10JFIF...

Now for the solution:

from io import BytesIO
from scipy import misc

numpyArray = misc.imread(BytesIO(content))

edited Jun 20 '20 at 09:12

Community

answered Apr 02 '19 at 13:40

Mark Setchell

why not just open the originl file with `misc.imread()` in the first place? – DrBwts Apr 02 '19 at 13:43
@DrBwts OP doesn't have the file, he gets its content from some web API - I think. – Mark Setchell Apr 02 '19 at 13:44

1 Answers1