Skip to content

Latest commit

 

History

History
68 lines (45 loc) · 2.32 KB

rough_alignment.rst

File metadata and controls

68 lines (45 loc) · 2.32 KB
Raw

Global 3D non-linear alignment

Global 3D non-linear alignment can be performed on a stack in chunks as well as the entire dataset (if all the serial sections are montaged and available). The following steps illustrate the global 3D non-linear alignment process.

Step 1 - Montage scapes generation

Montage scapes are downsampled versions of the serial sections and are used in the global 3D alignment process. Montage scapes can be generated as follows.

# Generate downsampled versions of montaged serial sections
python -m asap.materialize.render_downsample_sections --input_json <input_parameter_json_file> --output_json <output_json_file>

# Create a downsampled montage stack
python -m asap.dataimport.make_montage_scapes_stack --input_json <input_parameter_json_file> --output_json <output_json_file>

Step 2 - Generate tilepairs

3D tilepairs for the downsampled stack can be generated using the following command.

python -m asap.pointmatch.create_tilepairs --input_json <input_parameter_json_file> --output_json <output_json_file>

Step 3 - Generate point matches

3D point matches can be generated using the generated tile pairs. The following command can be used to generate point matches using a Spark cluster

python -m asap.pointmatch.generate_point_matches_using_spark --input_json <input_parameter_json_file> --output_json <output_json_file>

Step 4 - Solve for 3D non-linear transformations

This step in practice is done as a multi-step 3D alignment process, where a series of transformations (rigid, affine, non-linear) are computed and used as initialization for the computation of next higher order transformation.

A mesh based alignment can also be applied as a last step and is available in Bigfeta.

The command to run the solver is shown below.

python -m asap.solver.solve --input_json <input_parameter_json_file> --output_json <output_json_file>

NOTE: Each of the modules’ script include an example input json file for reference and the list of input and output parameters can also be listed using the –help option in each of the above commands.