Opening data and processing

Opening existing data

To work with data you recorded earlier, or data from another instrument, click Open Session in the main window.

1. Data folder: choose the folder that holds the session. QCLM scans it, and its subfolders named frames, cubes, results, capture, raw or data, then sorts what it finds:

   Found	Recognised by
   Raw frames	.raw, .bin, .dat, .hdr (ENVI), .tif / .tiff
   Spectral cubes	.npy, .h5 / .hdf5, .nc, .zarr folders
   Results	.json, .csv, .yml / .yaml
   White reference	file name contains whiteref or white_ref
   Dark reference	file name contains darkref or dark_ref

2. Choose the Experiment type: Hyperspectral or Sparse Wavelength.

3. Set the Processing options to apply on import: Normalize image, Savitzky-Golay destripe and ROI bounds.
4. Click Open Session ▶. QCLM uploads the data to the backend (Import to Server), parses it into a spectral cube and opens it in the acquisition window.

Supported file formats

Format	Extensions	Notes
ENVI	.hdr + data file	BSQ, BIL and BIP interleave; reads data type, byte order and wavelengths from the header
HDF5	.h5, .hdf5	Uses the first 3-D dataset; wavelengths are read from an attribute if present
NetCDF	.nc	Uses the first 3-D variable; wavelengths are read from a variable or attribute if present
NumPy	.npy	float32/float64; 3-D arrays in (bands, rows, cols) order, 2-D arrays as a single band
TIFF	.tif, .tiff	Multi-page TIFF, one page per band; 8/16-bit integer or 32/64-bit float
Zarr	.zarr	Float data (f4 / f8), uncompressed or zlib
Raw binary	.bin, .raw, .dat	QCLM's own format: a header of three little-endian ints (bands, rows, cols), then float64 values band by band

Processing options

The Processing tab controls how frames become a calibrated cube. The same options are available when importing a session.

Option	What it does
Reference correction	White/dark correction: converts raw counts into relative reflectance using a white reference and a dark reference. Set White ref and Dark ref.
Destriping	Removes column stripes caused by detector readout. Window and Order set the Savitzky-Golay smoothing.
Normalize	Rescales each band to the range 0–1.
Band cropping	Keeps only a range of bands (by band index) and drops the rest.
Downsample	Averages each n × n block of pixels into one, so the image shrinks by a factor of n in each direction.
Row / Column bounds	Crops the cube to a rectangular region of interest.

Click Process (Cube Build) to apply them. With Auto-process pipeline enabled, they are applied automatically after every scan.

Processing pipeline gives the exact formulas and the order of the steps.

Good references matter

Record the white reference on a uniform, highly reflective target, with the same settings as your sample. Record the dark reference with the source blocked. Correction divides by (white − dark), so noisy references add noise to every pixel.

Regions of interest and masks

The ROI tab defines which pixels the analysis uses.

• Draw ROI on canvas: draw a region on the image, or use Lasso for a freehand shape. Add current ROI stores it as a sub-ROI, and you can keep several.
• Measurement ROI: the region used for index plots and statistics.
• Mask tools: Draw and Erase with a Brush size, or Threshold mask to mask by intensity.
• Mask operations: Union, Intersect, Subtract and Invert mask. Clean mask removes small specks, and Remove borders clears the image edges.
• Mask I/O: Save to file and Load from file, so you can reuse masks across sessions.