Load, crop, and save a 10 bit HEIF

[kenwer]

I'm trying to load an input.HIF, crop it, and save output.HIF while keeping the format of the output image as close as possible to the format of the input image. In my case the input.HIF was created by a Sony ILCE-7M4. The camera seems to store the file with 10 bit luma & chroma, and uses 4:2:2 chroma sub sampling:

$ exiftool input.HIF | grep -Ei "Bit Depth|Chroma Format"
Chroma Format                   : 4:2:2
Bit Depth Luma                  : 10
Bit Depth Chroma                : 10

My initial approach was to use:

import pillow_heif
pillow_heif.register_heif_opener()
heif = pillow_heif.open_heif('input.HIF')
img = heif.to_pillow()
img.load() # required?
bit_depth = heif.info.get('bit_depth', 8)
chroma = heif.info.get('chroma', '420')
print(f"Attempting to preserve Bit Depth: {bit_depth} and Chroma Subsampling: {chroma}")
# imagine the crop here using PIL `img.crop()`
img.save('output.HIF', t="HEIF", quality=-1, bit_depth=bit_depth, chroma=chroma)

Which works but results in an output.HIF with only 8 bit:

$ python test.py                                          
Attempting to preserve Bit Depth: 10 and Chroma Subsampling: 422

$ exiftool output.HIF | grep -Ei "Bit Depth|Chroma Format"
Chroma Format                   : 4:2:2
Bit Depth Luma                  : 8
Bit Depth Chroma                : 8

I've also tried to use open_heif with convert_hdr_to_8bit=False and replaced line 3 in the snipped above like this:

heif = pillow_heif.open_heif('input.HIF', convert_hdr_to_8bit=False)

Wich results in heif.to_pillow() throwing a ValueError: unrecognized image mode. I suspect this indicates I cannot create a PIL image from it.

I could use heif.safe():

heif.save("output.HIF", format="HEIF", quality=-1, bit_depth=bit_depth, chroma=chroma)

But then I don't have a PIL image for easy cropping.

Any ideas on best practices how to load, crop & save such a HIF file while preserving the format?

If it matters, I'm using Python 3.13.7, and pillow_heif 1.1.0:

$ python --version
Python 3.13.7

$ pip list        
Package     Version
----------- -------
numpy       2.3.2
pillow      11.3.0
pillow_heif 1.1.0
pip         25.2

Any hints are appreciated.

[kenwer]

As a workaround to PIL's crop() for such 10 bit images, I'm using a numpy array (obtained via the array interface) that holds the cropped data and creating a new HeifImage using pillow_heif.from_bytes(). I also re-use the exif, xmp, and nclx_profile to keep as much meta data as I can. Here's the code:

def crop_hif(input_path, output_path, rect):
    heif = pillow_heif.open_heif(input_path, convert_hdr_to_8bit=False)
    np_array = np.asarray(heif[0])
    x, y, w, h = rect
    cropped_np_array = np_array[y:y+h, x:x+w]
    mode = heif[0].mode
    size = (cropped_np_array.shape[1], cropped_np_array.shape[0])
    data = cropped_np_array.tobytes()
    new_heif_image = pillow_heif.from_bytes(mode=mode, size=size, data=data)
    
    new_heif_image.save(output_path, 
                        format="HEIF", 
                        quality=-1, 
                        bit_depth=heif.info.get('bit_depth', 8), 
                        chroma=heif.info.get('chroma', '420'), 
                        nclx_profile=heif.info.get('nclx_profile'), 
                        exif=heif.info.get('exif'), 
                        xmp=heif.info.get('xmp'))

Are there better ways to handle/crop such HIFs? Any ideas are welcome.

[bigcat88]

The best way is to ask in the Pillow repo to prioritize developing 10/12 bit support (they have this issue)

[kenwer]

Oki, thanks and keep up the good work! : )