将base64的字符串转换回base64字节

Question

I've uploaded an image using OpenCV and then encode it with base64 encoding using base64 's b64encode .

>>> import cv2
>>> import base64
>>> image = cv2.cvtColor(cv2.imread("some_image.jpg"), cv2.COLOR_BGR2RGB)
>>> image_64 = base64.b64encode(image)
>>> image_64
b'//////////////////...
>>> type(image_64)
<class 'bytes'>

Then I convert it into a string using the str() method. This creates a string of the encoded image.

>>> image_64str = str(image_64)
>>> image_64str
b'//////////////////...
>>> type(image_64str)
<class 'str'>

Both of them (the <class 'bytes'> type and the <class 'str'> ) looks similar. I attempted to decode them using base64 's b64decode and the decode() function. However, an error occurred when I decoded the image_64str .

>>> image_64str.decode()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'str' object has no attribute 'decode'
>>> base64.b64decode(image_64str)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/Cellar/python/3.6.5/Frameworks/Python.framework/Versions/3.6/lib/python3.6/base64.py", line 87, in b64decode
    return binascii.a2b_base64(s)
binascii.Error: Incorrect padding

I fully understood what the errors were trying to tell me. But my question is, how can i convert the string of the encoded image ( image_64str ) back to bytes?

I've tried to use base64 's 'b64encode` again on the string. However, it returns an error.

>>> str_to_b64 = base64.b64encode(image_64str)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/Cellar/python/3.6.5/Frameworks/Python.framework/Versions/3.6/lib/python3.6/base64.py", line 58, in b64encode
    encoded = binascii.b2a_base64(s, newline=False)
TypeError: a bytes-like object is required, not 'str'

Please do tell if anybody noticed what i was missing. I am using Python 3.6. Thanks in advance.

EDIT: Adding more description to my question.

I was able to enable AWS API Gateway binary support. My purpose is to pass an image as binary data through a POST request to the API and convert that to a PIL object so that i can process it in the backend using AWS Lambda. With API Gateway, the binary data was encoded with base64 binary.

I opened the image as binary data using python's open function (there were two images that I wanted to pass through the API). Then i use i use a dictionary to hold both the two images binary data, like

data = {"data1": img_binary_data_1, "data2": img_binary_data_2}

I send the POST request using python request library. One of the argument that i can pass in the post function is data , so i passed the image data using that.

I was able to send the request. In the Lambda backend, I wanted to convert the binary data to a PIL object for further processing. However, it seems that the data was packed into JSON format and the base64 encoded binary image had been turned in to a python string. I confirmed this by printing the data in the log of AWS CloudWatch.

I tried to use .decode() , but base here you cannot decode a string.

I was able to decode the string using b64decode() , returning a byte object. However when try to convert it to a PIL Object like

img = imread(io.BytesIO(base64.b64decode(b64_string)))

I received an error saying

OSError: cannot identify image file <_io.BytesIO object at 0x1101dadb0>

I tried some of the solution from this link , but apparently you cannot do this with byte-object.

I have tried ti use PIL.frombuffer and PIL.frombytes . However, they returned the not enough data value when i am very sure about the sizer of the image (in this case (256, 256) ).

So my question is, how can i convert the base64 image in to a PIL object? I hope this helps to understand my question better. Thanks in advance.

Answer 1

Base64 is a binary -> char encoding so encoding an image makes sense, you get text bytes where a group of 6 bits is considered a character.

Now even if the above bytes are of characters, they are not python strings as python strings are utf-8.

When you convert the bytes to string it converts them to utf-8 and messes up the base64 padding (only = is allowed for padding) and what you get is a python string.

Now get the error when you decode it as it is not base64 encoding anymore. You can also not encode the string as base64 is bytes -> char and a string is not bytes.

Why are you converting the encoded bytes to string anyway? A little more description of your usecase would help.