Glossary

backend

A scientific processing library whose functions HTTomo can call, such as HTTomolibGPU, HTTomolib or TomoPy. See Processing libraries; developers should also read Integrating methods with httomo-backends.

block

The memory-sized portion of one process’s chunk passed through every method in a section before the next block is processed. See Blocks.

chunk

The portion of a section’s dataset assigned to one MPI process. See Chunks.

centre of rotation

The detector coordinate corresponding to the sample’s rotation axis. Accurate centring prevents characteristic reconstruction artefacts. See Centre of Rotation.

image key

A one-dimensional dataset that identifies each input frame as a projection, flat or dark image. See Darks and flats and Creating an NXtomo file.

intermediate dataset

A complete volume saved after a pipeline method, usually as an HDF5 file with the main array at /data. See Run output, logs and monitoring.

loader

The first pipeline method, responsible for locating input arrays and presenting projection data and auxiliary information to HTTomo. See Standard tomography loader.

method

One loader, processing or output operation configured as an entry in a pipeline. See Pipeline file reference for the fields that define a method entry and Available methods for supported methods.

method template

A YAML description of a supported method, its parameters and defaults, generated from httomo-backends metadata. See YAML templates and Available methods.

monitor

Optional runtime instrumentation that reports aggregate or block-level timings. See Run output, logs and monitoring and Command-line interface.

NXtomo

The NeXus application definition for tomography data. An NXtomo entry links projection, image-key and rotation-angle datasets in a standard hierarchy that HTTomo can discover automatically. See Creating an NXtomo file.

padding

Extra neighbouring slices supplied to a method so that operations near a block boundary have sufficient context. See Padding.

parameter sweep

Repeated execution of a method for several candidate parameter values, with images saved for comparison. See Parameter Sweeping.

pattern

The orientation in which methods consume data, principally projection or sinogram order. Pattern changes determine section boundaries and can require a re-slice; see Sections and Re-slicing.

pipeline

The ordered sequence of methods that HTTomo executes. Older material may call this a process list. See Pipelines, Pipeline file reference and Ready-to-use pipelines.

preview

A loader selection that crops the angular or detector dimensions before processing. See Previewing.

rank

The identifier of one MPI process participating in a parallel run. Each rank normally receives one chunk; see Chunks and HTTomo execution from a YAML pipeline to output. The pipeline is divided into sections; within each section, data is distributed into chunks and processed block by block..

re-slice

Redistribution or transposition of data when consecutive sections use different processing patterns. See Re-slicing.

section

A consecutive group of compatible methods that use the same processing pattern and are sized and executed together. See Sections and HTTomo execution from a YAML pipeline to output. The pipeline is divided into sections; within each section, data is distributed into chunks and processed block by block..

side output

A named value produced alongside the main dataset and referenced by a later pipeline method. See Side outputs for syntax and examples.

wrapper

HTTomo’s adapter between a backend function and the common pipeline execution interface. See Method wrappers and Architecture and internals.