How to convert an imagehash to a numpy array?

Question

I use python / imagehash.phash to search for photos similar to a given one, i.e. photos so that the hamming distance is zero. Now I also would like to detect photos that are flipped or rotated copies - the hamming distance is > 0. Instead of flipping or rotating the photo and then calculating an new pHash I would like to derive the pHash of a rotated / flipped photo directly from the original pHash.

My idea is to convert the hash into an numpy array, use np.flip etc and convert the array back into a pHash.

Questions:

1. How can I convert the hash into an numpy array and back?
2. is there a better way to search for rotated / flipped images given the hash of the original photo?

Answer 1

From reading the source code, I found that you can convert between an imagehash.ImageHash object and a binary array using .hash and the imagehash.ImageHash() constructor.

from PIL import Image
import imagehash

img = Image.open("house.jpg")
image_hash_obj = imagehash.phash(img)
print(image_hash_obj)
hash_array = image_hash_obj.hash
print(hash_array)
hash_array[0, 0] = False
image_hash_obj2 = imagehash.ImageHash(hash_array)
print(image_hash_obj2)

Output:

fde1921f6c20b34c
[[ True  True  True  True  True  True False  True]
 [ True  True  True False False False False  True]
 [ True False False  True False False  True False]
 [False False False  True  True  True  True  True]
 [False  True  True False  True  True False False]
 [False False  True False False False False False]
 [ True False  True  True False False  True  True]
 [False  True False False  True  True False False]]
7de1921f6c20b34c