Saving images

Current limitations of encoder

  • Libheif does not support editing files in place(it is not possible just to change metadata)

  • HEIF format does not provide information in what quality image was encoded

Save parameters

Method save in both HeifFile and Pillow supports the same parameters. There is only two differences between them:

  • Pillow’s save by default has save_all=False and HeifFile has save_all=True

  • When saving to memory for HeifFile you do not need to specify format parameter.

Here is description of it: save()

Saving 16 bit images

All 16 bit images that was created with:

Will be saved by default in 10 bit mode.

To save 16 bit image in 12 bit set options().save_to_12bit to True.

Images opened from file will be saved based on info["bit_depth"] value if it is present.

Order Of Images

All information here is only for files that has multiply images and when first image in file is not a PrimaryImage

Lets imagine that we have file with 3 images and second image in file is a primary image.

im = open_heif("image.heif")
im_pil = Image.open("image.heif")

Both im and im_pil points to the second image in file, that is a primary image.

This asserts will pass:

assert im.info["primary"] == True
assert im[0].info["primary"] == False
assert im[1].info["primary"] == True
assert im[2].info["primary"] == False

assert im_pil.info["primary"] == True
im_pil.seek(0)
assert im_pil.info["primary"] == False
im_pil.seek(1)
assert im_pil.info["primary"] == True
im_pil.seek(2)
assert im_pil.info["primary"] == False

And next code will produce the same behavior results(during enumerating all frames):

for frame in im:
    print(frame.mode, frame.size)

for frame in ImageSequence.Iterator(im_pil):
    print(frame.mode, frame.size)

Primary image when saving image sequences

First image will be primary, unless after first image there is no images with info[“primary”] set to True. So, last image with info[“primary”] == True will be the primary image.

In most cases you do not need to bother yourself with this, as other image formats are much simpler and has no such abilities.

When you need to change primary image for some reason, just set info[“primary”] of image to True and all images after that should not have info[“primary”] == True.

Method save supports primary_index parameter, that accepts index of image or -1 to set last image as PrimaryImage.

Specifying primary_index during save has highest priority.

NCLX color profile

By default, since version 0.14.0, if the image already had an NCLX color profile, it will be saved during encoding.

Note

If you need old behaviour and for some reason do not need NCLX profile be saved you can set global option SAVE_NCLX_PROFILE to False.

To change it, you can specify your values for NCLX color conversion for save operation in two ways.

Set output NCLX profile:

Note

Avalaible only from 0.14.0 version.

buf = BytesIO()
im.save(buf, format="HEIF", matrix_coefficients=0, color_primaries=1)

In this case the default output NCLX profile will be created, and values you provide in such way, will replace the values from default output profile.

Edit NCLX profile in image.info:

buf = BytesIO()
im.info["nclx_profile"]["matrix_coefficients"] = 0  # code assumes that image has already "nclx_profile"
im.info["nclx_profile"]["color_primaries"] = 1
im.save(buf, format="HEIF")

Under the hood it is much complex, as second way will change the input NCLX profile.

The preferable way is to specify new NCLX values during save.

Here is additional info, from the libheif repo with relevant information:

https://github.com/strukturag/libheif/discussions/931 https://github.com/strukturag/libheif/issues/995

Lossless encoding

Note

Parameter matrix_coefficients avalaible only from 0.14.0 version.

Although the HEIF format is not intended for lossless encoding, it is possible with some encoders that support it.

You need to specify matrix_coefficients=0 (which will tell libheif to perform the conversion in the RGB color space) and chrome subsampling equal to “4:4:4”.

im_rgb = Image.merge(
    "RGB",
    [
        Image.linear_gradient(mode="L"),
        Image.linear_gradient(mode="L").transpose(Image.ROTATE_90),
        Image.linear_gradient(mode="L").transpose(Image.ROTATE_180),
    ],
)
buf = BytesIO()
im_rgb.save(buf, format="HEIF", quality=-1, chroma=444, matrix_coefficients=0)

That’s all.