Reconstruction Parameters

This page summarises the parameter tabs on the Main tab row. All fields map directly to TomoCuPy command-line flags; refer to the TomoCuPy docs for deeper coverage.

Recon tab

Reconstruction algorithm

  • FBP — direct filtered backprojection (fastest, most common)

  • gridrec — grid-based FBP, often good for 360° scans

  • LPREC — log-polar; can be faster on very large sinograms

nsino-per-chunk

Number of sinograms processed per chunk. Increase for faster reconstructions if GPU memory allows; decrease if you see OOM errors.

binning

Downsampling factor. 0 = none, 1 = ½, 2 = ¼, etc.

start-row / end-row

Slice range for Try (full volume is always used for Full unless otherwise specified).

Hardening tab

Beam hardening correction parameters. Leave at default unless you have a calibrated hardening model.

Phase tab

Phase retrieval:

  • method — Paganin / Bronnikov / CTF

  • alpha — regularisation strength

  • energy — beam energy (keV)

  • pixel size, distance — sample-to-detector geometry

Rings tab

Ring artifact removal:

  • algorithm — fw (Fourier-Wavelet), none, etc.

  • sigma — smoothing strength

  • level — wavelet level

Geometry tab

Rotation axis geometry flags (tilt, lateral shift).

Data tab

Projection data preprocessing: flat / dark correction, projection range, bright / dark frame source, etc.

Performance tab

Threading, blocked-views, chunking, and other performance knobs.

Per-dataset persistence

When you change a parameter, its value is written to a JSON sidecar next to the projection file. Re-selecting the file later restores the last values. The Batch tab’s Apply parameters to selected action copies a sidecar from one row to all checked rows.

See Reconstruction Workflow for the end-to-end workflow.