Would like to know little bit insight why one image size is increased and other one decreased?

Hello there, I have downloaded two .tiff images from cytomine website:

1. https://cytomine.com/collection/hsm0170/400-human-ileons (it’s a huge image 158 GB )
2. https://cytomine.com/collection/uliege001/cat-rectum (size 847 Mb) After that, I converted them into ome.tiff using the below code (bcoz bioformat was unable to read these image resolution layers).

def tiles(stack, tileshape):
    for y in range(0, stack.shape[0], tileshape[0]):
        for x in range(0, stack.shape[1], tileshape[1]):
            yield stack[y: y + tileshape[0], x: x + tileshape[1]]


with tifffile.imread(input_path, aszarr=True) as store:
    group = zarr.open(store, mode='a')
    assert isinstance(group, zarr.Group)
    with tifffile.TiffWriter(output_path + f'HSM0170.ome..tiff', bigtiff=True) as tiff:
        start_time = timeit.default_timer()
        for r in group.keys():
            stack = group[r]
            width = stack.shape[0]
            length = stack.shape[1]
            # assert stack.ndim == 3  # YXC
            a = tiles(stack, (stack.chunks[0], stack.chunks[1]))

            options = {'tile': (stack.chunks[0], stack.chunks[1]), 'compression': 'JPEG'}
            tiff.write(tiles(stack, (stack.chunks[0], stack.chunks[1])), shape=stack.shape,
                       dtype=stack.dtype, **options)
    end_time = timeit.default_timer()
    total_time = float(end_time - start_time)
    print(total_time)

It got converted and I receive generated image first image size was increased to 247.2 GB and the second image size was decreased to 713.5 MB. I want to know a little bit of insight detail why the first image size is increased and the second image size is decreased while applying the same code. Which parameter is crucial while converting, is it jpeg quality?

[cgohlke]

For pyramidal OME-TIFF you have to reserve subifds for all subresolutions when writing the base/first level.

This should work, although it will probably take several hours to decode and compress several terabytes of image data:

import tifffile
import zarr

input_path = 'HSM0170-20x20.tiff'
output_path = 'HSM0170.ome.tif'


def tiles(stack, tileshape):
    for y in range(0, stack.shape[0], tileshape[0]):
        for x in range(0, stack.shape[1], tileshape[1]):
            yield stack[y : y + tileshape[0], x : x + tileshape[1]]


with tifffile.Timer():

    with tifffile.imread(input_path, aszarr=True) as store:
        group = zarr.open(store, mode='r')
        assert isinstance(group, zarr.Group)

        with tifffile.TiffWriter(output_path, bigtiff=True) as tiff:
            for i, r in enumerate(tifffile.natural_sorted(group.keys())):
                stack = group[r]
                assert stack.ndim == 3  # YXC
                options = {
                    'tile': (stack.chunks[0], stack.chunks[1]),
                    'compression': ('jpeg', 70),
                }
                if i == 0:
                    options['subifds'] = len(group.keys()) - 1
                else:
                    options['subfiletype'] = 1
                tiff.write(
                    tiles(stack, (stack.chunks[0], stack.chunks[1])),
                    shape=stack.shape,
                    dtype=stack.dtype,
                    **options,
                )

[cgohlke]

There was another issue in the code. You need to iterate over the zarr groups in the correct order for i, r in enumerate(tifffile.natural_sorted(group.keys())):.