Reversing bits of Python Numpy Array

Question

Suppose I have a numpy array of dtype uint16, how do I efficiently manipulate each element of the array so that the bits are reversed?

eg. 0001111010011100 -> 0011100101111000

The existing solutions on this website seem to suggest printing the number into a string which will be really slow for arrays.

Example of what I want to do:

test = np.array([128, 64, 32, 16, 8, 4, 2, 1]).astype(np.uint16)
out = reverse_bits(test)
print(out)
>> array([  256,   512,  1024,  2048,  4096,  8192, 16384, 32768], dtype=uint16)

I have an image mask saved as a 16bit tiff where the bits are coded to represent different classes. I just wanted a bijective way to visualise an image mask and "amplify" values with small differences.it wasn't strictly necessary as I am aware of alternatives. — matohak
– matohak, Commented Feb 5, 2020 at 2:38

user12734467 · Accepted Answer · 2020-02-04 16:36:30Z

2

This will reverse bits in each element of an array.

def reverse_bits(x):

    x = np.array(x)
    n_bits = x.dtype.itemsize * 8

    x_reversed = np.zeros_like(x)
    for i in range(n_bits):
        x_reversed = (x_reversed << 1) | x & 1
        x >>= 1
    return x_reversed

answered Feb 4, 2020 at 16:36

user12734467

Sign up to request clarification or add additional context in comments.

1 Comment

fuzzyTew Over a year ago

This solves the problem but does not appear efficient. Other options may include packbits at least.

Rgaddi · Accepted Answer · 2021-08-20 17:56:09Z

2

Here's one based off some old HAKMEM tricks.

def bitreverse16(x):
    x = ((x & 0x00FF) << 8) | ((x & 0xFF00) >> 8)
    x = ((x & 0x0F0F) << 4) | ((x & 0xF0F0) >> 4)
    x = ((x & 0x3333) << 2) | ((x & 0xCCCC) >> 2)
    x = ((x & 0x5555) << 1) | ((x & 0xAAAA) >> 1)
    return x

answered Aug 20, 2021 at 17:56

Rgaddi

4551 gold badge4 silver badges11 bronze badges

Comments

DStauffman · Accepted Answer · 2023-05-30 23:06:23Z

2

How about this version. I timed it about 6X faster than the bitreverse16 and 20X faster than the reverse_bits function for the given example vector.

def reverse_bits2(x):
    dtype = np.asanyarray(x).dtype
    return np.flip(np.packbits(np.flip(np.unpackbits(x.view(np.uint8)))).view(dtype))

answered May 30, 2023 at 23:06

DStauffman

4,4492 gold badges23 silver badges32 bronze badges

Collectives™ on Stack Overflow

Reversing bits of Python Numpy Array

3 Answers 3

1 Comment

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

1 Comment

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related