I have a hexadecimal string "89-50-4E-47-0D-0A-1A-0A-00-00-00-0D-49" to be specific this will contain the data of an image.
I want to convert it to a Numpy array or possibly reconstruct an OpenCV image from the said data.
The width and height will also be supplied so the dimensions of the Numpy array is known.
How can I construct a Numpy array from the above string?
We could use np.fromiter, and cast the individual strings to hexadecimal, using the base argument in int, and then to integer type using the dtype argument in np.fromiter:
s = "89-50-4E-47-0D-0A-1A-0A-00-00-00-0D-49"
np.fromiter((int(x, 16) for x in s.split('-')), dtype=np.int32)
# array([137, 80, 78, 71, 13, 10, 26, 10, 0, 0, 0, 13, 73])
You can use list comprehension and the built-in int module to convert from hexadecimal to decimal the splitted string:
import numpy as np
hex_string = '89-50-4E-47-0D-0A-1A-0A-00-00-00-0D-49'
np.array([int(x, base=16) for x in hex_string.split('-')])
import numpy as np
arr = np.array([int(x, 16) for x in "89-50-4E-47-0D-0A-1A-0A-00-00-00-0D-49".split("-")])
print(arr)
You can split the string by dash and convert individual base-16 numbers into int.
>>> import numpy as np
>>> hext_str = "89-50-4E-47-0D-0A-1A-0A-00-00-00-0D-49"
>>> np.array([int(x, 16) for x in hex_str.split("-")])
array([137, 80, 78, 71, 13, 10, 26, 10, 0, 0, 0, 13, 73])
Provided that (n, m) are dimensions of your image you can use on the result the .reshape((n, m)) method of np.array.
If memory efficiency is of concern two digits in a hexadecimal number corresponds to an unsigned 8-bit integer (ie, numbers between 0 and 255).
To return to the original string you can use format(number, '02x') (zero padded 2-length hexadecimal number string)
hex_str = "89-50-4E-47-0D-0A-1A-0A-00-00-00-0D-49"
arr = np.fromiter((int(x, 16) for x in hex_str.split('-')), dtype=np.uint8)
# array([137, 80, 78, 71, 13, 10, 26, 10, 0, 0, 0, 13, 73],
dtype=uint8)
The array would only take up 13 bytes of space, as opposed to the default array type for integers (np.int64) which would take up 104 bytes.
Once could return it to its original string form in the following way:
hex_str = '-'.join((format(x, '02x') for x in arr)).upper()
It is possible to use bytes.fromhex along with np.frombuffer
import numpy as np
hex_string = "ff01"
arr = np.frombuffer(bytes.fromhex(hex_string), dtype=np.uint8)
# array([255, 1], dtype=uint8)
in your case
import numpy as np
hex_string = "89-50-4E-47-0D-0A-1A-0A-00-00-00-0D-49"
arr = np.frombuffer(bytes.fromhex(hex_string.replace("-", "")), dtype=np.uint8)
np.array([int(x, 16) for x in "89-50-4E-47-0D-0A-1A-0A-00-00-00-0D-49".split("-")])