Darksift Master FITS

Prepare master-frame files from your own same-grid FITS images on iPhone and iPad. Darksift combines a selected batch into a real 32-bit floating-point FITS master. Compare a source, the master and a count map of excluded samples. Keep originals, settings and notes together for the next session. A NUMERIC WORKFLOW Choose Mean, Median or MAD mean. Mean and Median exclude no samples. MAD mean is one pass: values farther than your threshold × 1.4826 × median absolute deviation are excluded before an unweighted mean. The absolute rejection floor is 0.000001. Inspect exact source and result values at a one-based pixel coordinate. CONTROL THE INPUTS Import through native Files. Include or exclude a frame, remove it, add more or share an unchanged original. All included images must have identical dimensions. Known exposure, gain, temperature, instrument, filter, Bayer pattern, binning, image type and unit disagreements stop processing. If headers omit these fields, explicitly confirm that you have checked the sensor and settings. Matching headers do not prove physical camera compatibility. KEEP YOUR WORK Save twelve local batches, up to thirty-two retained originals each. Rename, duplicate or delete batches; save method, threshold, include flags and notes. Reopen a batch and rebuild from its retained source files. Production starts empty; an explicitly labelled original synthetic sensor study is available. SHARE REAL RESULTS Write Master.fits, a full Rejected.fits count map, a two-page PDF preview, input summaries with original SHA256 hashes, sparse rejection CSV, recipe JSON and readme, together in a ZIP. The CSV lists at most 100,000 nonzero rejection pixels; the FITS map remains complete. Preview the PDF with Quick Look and share actual files using the system sheet. Originals are shareable individually and are not included in the derived ZIP. Recipe JSON documents parameters, not an importable backup. VALUES STAY UNSCALED Preview uses a percentile display stretch. It is never applied to FITS output. No normalization, registration, debayering, light-frame calibration or star alignment is performed. The master keeps only common validated acquisition fields, not arbitrary WCS or other headers. Spatial sigma describes the image and is not a measured noise-reduction guarantee. Masters need testing in your own compatible downstream workflow. INPUT LIMITS Uncompressed single primary-HDU 2D FITS; BITPIX 8, 16, 32, −32 or −64, finite BSCALE/BZERO and finite values representable in float32. Up to 8 million pixels per image, 70 MB per file and 64 million decoded values per batch (256 MB). Include 2–32 frames. BLANK, random groups, extra HDUs, cubes, compressed FITS, XISF, JPEG and camera RAW are unsupported. This is intended for same-grid dark/bias preparation, not normalized flats or aligned night-sky images. One purchase. No subscription, account, advertising, analytics, camera capture or telescope connection. Files stay on your device unless you choose to share them.