First release. Pure-R reader for Diaspective Vision TIVITA recordings — no vendor SDK, no compiled code.
tivis_read_cube() reads a *_SpecCube.dat container and returns a
(rows, cols, bands) array with a wavelengths attribute, matching the
convention used by hyperspectR::hsi_cube(). Supports reading a subset of
bands.tivis_read_header() decodes the 12-byte dimension header without touching
the samples, so large archives can be scanned cheaply. It validates the file
size against the header and errors on a mismatch rather than returning a
silently recycled array.tivis_get_metadata() parses the Suite's *_meta.log: an INI dialect with
German section names, latin1 encoding and comma decimal separators. Values
are coerced to numerics and logicals where appropriate.tivis_reference_images() locates the parameter maps the Suite exports
beside each capture (RGB, oxygenation, NIR perfusion, THI, TWI, OHI).tivis_list_measurements() walks an archive and returns one row per
recording.tivis_measurement() bundles cube, wavelengths, metadata, references and
timestamp for a single capture.The container is undocumented by the vendor. The layout implemented here —
big-endian float32, band-interleaved-by-pixel, pixels column-major with
y varying fastest, behind three big-endian uint32 dimensions, on a
500–995 nm axis in 5 nm steps — was established by measuring the sample
stream's own periodicity.
Autocorrelation of the raw stream peaks at lag 100 (r = 0.95), the band count, with harmonics at 200/300/400: band interleaving by pixel. Autocorrelation of an extracted band plane peaks at lag 480 (r = 0.99), the image height, versus r = 0.43 at lag 640, the width: pixels run down columns.
During development this was briefly documented as row-major on the strength of a correlation against the Suite's RGB export. That was incorrect — the comparison was made on downsampled images, where the tiling artefact it should have exposed was averaged away. The vendor PNG is rotated and rescaled relative to the cube and is not usable as pixel-level ground truth.