sam

A library to split and merge high-resolution 3D images. Currently supports nii format only.

prerequisites

• Python 2.7.5
• Numpy 1.12.1
• Nibabel

how to use

Split:

naive split strategy

import imageutils as iu
img = iu.ImageUtils(filepath="/path/to/image.nii")
img.split(first_dim=5, second_dim=5, third_dim=5, local_dir="/path/to/output_dir", filename_prefix="bigbrain")

multiple writes strategy

# mem = 12g
import imageutils as iu
img = iu.ImageUtils(filepath="/path/to/image.nii")
img.split_multiple_writes(Y_splits=5, Z_splits=5, X_splits=5, out_dir="/path/to/output_dir", mem=12*1024**3, filename_prefix="bigbrain",
                          extension="nii")

Merge:

buffered slices strategy:

import imageutils as iu
import numpy as np
img = iu.ImageUtils(reconstructed="/path/to/img.nii", first_dim=3850, second_dim=3025, third_dim=3500, dtype=np.uint16)
img.reconstruct_img(legend="/path/to/legend", "clustered", mem=0)

buffered blocks strategy:

• cluster reads strategy

# mem=12g
import imageutils as iu
import numpy as np
img = iu.ImageUtils(reconstructed="/path/to/img.nii", first_dim=3850, second_dim=3025, third_dim=3500, dtype=np.uint16)
img.reconstruct_img(legend="/path/to/legend", "clustered", mem=12*1024**3)

• multiple reads strategy

# mem=12g
import imageutils as iu
import numpy as np
img = iu.ImageUtils(reconstructed="/path/to/img.nii", first_dim=3850, second_dim=3025, third_dim=3500, dtype=np.uint16)
img.reconstruct_img(legend="/path/to/legend", "multiple", mem=12*1024**3)

License

This project is licensed under the MIT License - see the LICENSE file for details