channel

[thewtex]

• ENH: tczyx, etc. support
• PERF: itkwasm downsampling in zyx order

[will-moore]

Thanks @thewtex,

I tried downsampling an image that has C dimension, hoping that the 2 separate channels would be downsampled independently, but it seems that the first channel is incorrectly resampled since it ends up being black:

import ngff_zarr as nz
import numpy as np

# 4D data
data = np.random.randint(0, 256, 16777216, dtype=np.uint8).reshape((2, 128, 256, 256))
image = nz.to_ngff_image(data, dims=['c', 'z', 'y', 'x'])

# downsample
multiscales = nz.to_multiscales(image, scale_factors=[2], chunks=64)

print("multiscales.images", multiscales.images)

# check that both channels have pixel data
for i, img in enumerate(multiscales.images):
    print("resolution", i)
    print("  ch0 max", img.data[0,:,:,:].max().compute())
    print("  ch1 max", img.data[1,:,:,:].max().compute())

This prints the following, showing that the first channel is black for the downsampled resolution:

resolution 0
  ch0 max 255
  ch1 max 255
resolution 1
  ch0 max 0
  ch1 max 211

[thewtex]

@will-moore thanks for testing! I pushed a fix for that.

[will-moore]

Working for me now, thanks!