-
Notifications
You must be signed in to change notification settings - Fork 12
N5 exporter details
The number of scale levels is determined by the image size and specified block size. The exporter will downsample an image only if the block size is striclty smaller than the image size in every dimension.
If image size is 100 x 100 x 100 and the block size is 100 x 100 x 100
will write one scale level because the whole image fits into one block
at the first scale level.
If image size is 100 x 100 x 100 and the block size is 64 x 64 x 64
will write two scale levels: The first scale level will have a 2 x 2 x 2 grid
of blocks.
If image size is 100 x 100 x 32 and the block size is 64 x 64 x 64
will write one scale level because the third dimension of the image is
smaller than the third dimension of the block.
The N5 exporter always downsamples images by factors of two in all dimensions. There are two downsampling methods:
N5 will take even-indexed samples and discard odd-indexed samples along each dimension.
N5 will average adjacent samples along each dimension. This results in a "half-pixel" shift, which will be reflected in the metadata.
If the overwrite option is not selected in the dialog, the exporter will determine if the
write operation would overwrite or invalidate existing data. If so, it prompts the user with a
warning dialog, asking if data should be overwritten.
Data could be overwritten if:
- the path where data will be written already exists and contains data.
- a parent of the path where data will be written already exists and contains data.
- here, the newly written data would be inaccessible, because data arrays must be leafs of the hierarchy tree.
If the overwrite option is selected in the initial dialog, the user will not be prompted, but
data will be overwritten if needed as explained below.
A dataset exists at /image. User attempts to write data into /image. This warns the user
about overwriting because an array already exists at that location.
A dataset exists at /data/s0. User attempts to write data into /data using N5Viewer metadata.
This warns the user about overwriting because when writing N5Viewer metadata, the plugin will
write the full resolution level of the multiscale pyramid at location /data/s0, but an array
already exists at that location.
If the user decides to overwite data the N5 exporter will completely (array data and metadata) remove any groups that cause conflicts before writing the new data.
- If a dataset already exists at a path where new data will be written, then all data at that path will be removed.
- If a dataset already exists at a parent path where new data will be written, then all data at that parent path will be removed.
A dataset exists at /image. User attempts to write data into /image/channel0. This warns
the user about overwriting because the newly written data would be a child path of existing
data, and therefore be invalid. If the user decides to overwrite, all data at /image will be
removed before writing the new data to /image/channel0.