Standalone merge

Use merge_tile_labels when you already have per-tile labels from your own pipeline and only need the boundary-stitching step.

From a dask array

import dask.array as da
import numpy as np
from blockbuster import merge_tile_labels

# Your own tiling + segmentation
image = da.from_zarr("image.zarr").rechunk((1, 1024, 1024))

def my_fn(tile: np.ndarray) -> np.ndarray:
    ...  # your segmentation code
    return labels.astype("int32")

labeled = image.map_blocks(
    my_fn, dtype="int32",
    meta=np.empty((0,) * image.ndim, dtype="int32"),
)

# Merge: boundary scan → connected components → relabel
merged = merge_tile_labels(labeled, write_to="labels.zarr", progress=True)

From a pre-written zarr

If your pipeline already wrote per-tile labels to zarr:

from blockbuster import merge_tile_labels

merged = merge_tile_labels(
    "my_staged_labels.zarr",
    input_component="raw_labels",  # component name inside the zarr
    write_to="merged.zarr",
    sequential_labels=True,        # renumber to 1..N
)

With overlap halos

If your tiles were computed with da.overlap.overlap(depth=20), trim the halos before merging:

from blockbuster import merge_tile_labels

# labeled was computed with da.overlap.overlap(depth=20)
merged = merge_tile_labels(labeled, write_to="labels.zarr", overlap=20)

Integration with other frameworks

blockbuster's merge step is framework-agnostic. Any pipeline that produces a dask array of integer labels (one value per voxel, distinct per tile) can use merge_tile_labels:

# zarr → your pipeline → per-tile labels
labeled = your_pipeline(da.from_zarr("image.zarr"))  # dask.array.Array

from blockbuster import merge_tile_labels
merged = merge_tile_labels(labeled, write_to="final.zarr")