Sony ARWs of cameras without measured colour matrices were edited on the
relative scale (6500 K / 0 = as shot). Their SR2SubIFD stores the
multipliers of the camera's presets at 2500, 3200, 4500, 6000 and 8500 K;
lightcraft_raw::kelvin_presets reads the as-shot multipliers on the curve
through them and they take the spectral matrices' place in the #730
design (as-shot Kelvin and tint only; look, Bradford adaptation and the
pixel-identical legacy upgrade unchanged).
Covered: ILCA-99M2, ILCA-77M2, SLT-A33, DSC-RX10, ILCE-3500, ILCE-7S.
Kelvin-dial shots read back 3400 -> 3396, 5200 -> 5193, 4700 -> 4708,
5600 -> 5594 K; Shade 7493 K. Off-curve constants (isotherm tilt 0.74,
tint spacing 1) measured from the 14 Sony bodies with both spectral
matrices and presets.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Masked lines are measured against the outermost line's level when it sits
just above BlackPoint (K-70, K-1), saturated outer lines count when dark
ones follow (645Z), and a frame whose interior is within the tolerance
keeps its whole area. Line means read samples with get(). Render cache
version 62: PEF frames change size.
The DNG CFAPattern describes the top-left of the ActiveArea, but RawImage::cfa is sensor-anchored (the crop to the active area shifts it by the area's offset). The DNG reader now moves the tag's pattern out to the sensor origin, and the DNG writer states the pattern back at the ActiveArea. A DNG whose area starts on an odd row or column (the Canon EOS 60D DNG of #1086, top 51) no longer reads RGGB as GBRG.
Cause: a floating-point CFA DNG is read as stored (no check of its float
samples). AHD and PPG clamp each green estimate between its two neighbouring
greens with `est.clamp(a.min(b), a.max(b))`; when both greens are NaN the
bounds are NaN and `f32::clamp` panics (ppg.rs:45, ahd.rs green pass). Any
float DNG with two adjacent NaN greens crashed decode + develop. Infinities
alone did not panic but turned into NaN in the demosaic arithmetic.
Fix, at two layers:
- demosaic: a `clamp_between` helper replaces the three clamps. With finite
or infinite bounds it is the same `f32::clamp` call (bit-identical
results); with NaN bounds it returns the estimate unclamped, so the panic
is gone whatever puts NaN into the mosaic.
- DNG decode: non-finite float samples are replaced where the white level is
known. NaN and -inf read as black (0), +inf as the white level (a clipped
highlight). No sensor records NaN or infinity, so a valid file is never
touched; finite samples, negative or above white included, stay as stored.
Tests: `demosaic::tests::non_finite_samples_do_not_panic` (NaN block, +/-inf
and an all-NaN mosaic through Bilinear, PPG and AHD; plus `clamp_between`
equal to the old clamp bit for bit over finite, signed-zero and infinite
values) and `non_finite_float_samples_decode_and_develop` (a float CFA DNG,
uncompressed and Deflate, with NaN and +/-inf samples: decoded values
checked, develop with all three methods gives a finite image). Both panicked
before the fix at ppg.rs:45 / ahd.rs.
Co-Authored-By: Claude <noreply@anthropic.com>
Version 66 keeps version 64's as-shot white balance and black level
offsets; its word 661 is below where the sensor clips, so the white
level stays measured from the data. Checked on 33 R6 Mark III files
from 12 shoots: all now open with their as-shot white balance, black
levels match the masked columns.
Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017JKdR4oZ3jRSPPwMLLTa2U
New export format `rendereddng` (export dialog: Format DNG, "DNG of: Rendered
photo"; `app.export`; MCP export tool): the rendered photo, with the edit,
crop, size, output sharpening and watermark baked in, written by the new
`lightcraft_raw::write_rendered_dng` as a LinearRaw DNG of 16-bit float Deflate
tiles in the export colour space. ColorMatrix1 (XYZ -> space), ForwardMatrix1
(space -> XYZ D50), a neutral AsShotNeutral and an identity ProfileToneCurve
make raw converters show it as rendered. HDR edits keep highlights above white.
Bad input (wrong sample count, zero or huge sizes, NaN/inf, out of half-float
range) gives an error or is clamped, never a panic.
Also fixes the Rust 1.99 clippy lint chunks_exact_to_as_chunks in
pipeline/spots.rs that fails `cargo xtask ci` on main.
Sources: Adobe DNG Specification 1.4+ (LinearRaw, ColorMatrix/ForwardMatrix,
ProfileToneCurve); own design.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
On merge: the quick preview path orients like the whole-file path (preview_orientation with the
raw's header prefix, so a portrait raw whose embedded JPEG carries no orientation is not sideways),
and its doc links point at preview_window.
`lightcraft_raw::sparse::header_plan` says which byte ranges of a raw file the header
probe (`probe_info`, `lightcraft_meta::extract`) reads, given a buffer of the file's length
in which only some ranges are loaded: everything but the bulk sample data (large TIFF
strips/tiles minus the first chunk's coding headers, a CR3's image tracks minus the raw
track's CRX headers and QP maps, a RAF's CFA strip, an RW2's raw data). `preview_plan`
also reads what `embedded_preview` needs, for raws shown from their preview. NEF, ORF and
PEF (their header probe looks at samples) and anything else are read whole.
The corpus test checks, for every raw.pixls.us sample the plan describes, that probing the
partly loaded buffer gives exactly what the whole file gives, with the skipped bytes zero and
again 0xFF: CR3 2.7 %, DNG 2.4 %, CR2 3 %, RAF 6 %, RW2 7 %, ARW 8 % of their bytes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
HsvTable::apply took the value through linear -> sRGB -> linear even when the table's value
scale is 1 (every hue/saturation-only table, as fitted camera looks are), and wrapped hues with
rem_euclid (an fmodf call) three times per pixel. The value is now kept as it is for a unit
scale and the sRGB-encoded value is reused otherwise; hues are wrapped by a helper that equals
rem_euclid bit for bit (tested) without the call for values within one period.
2048 px render of a 24 MP ARW, single thread: 16.4 -> 15.1 G instructions, 4.94 -> 4.76 G cycles.
Sources: own design.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The CRX 0x200 QP map opens with its first symbol as a 32-bit literal; every
later symbol is Rice coded with a limited-length escape: 23 zeros, the
terminating 1 and the symbol in 8 bits (the ITU-T T.87 layout for a 32-bit
limit and 8-bit symbols, without its minus-one offset). The old rule (an
escape after 16 zeros with a 16-bit literal) matched only the first symbol,
which is all the R8 and R7 maps need. R5 Mark II maps (QP 150-162) jump by up
to 12 where the Rice parameter is 0, giving 18- to 22-zero prefixes that
misparsed ("Canon CRX adaptive QP value -21087"): 174 of 306 sampled files
from a private R5 Mark II library opened as previews. Every QP map of its
33,665 C-RAW files now decodes to the exact end of its entropy stream; the six
oracle fixtures are unchanged. Established by black-box analysis of the
bitstream; no third-party decoder source consulted.
Adds the four CC0 R5 Mark II samples (raw.pixls.us 7881-7884) to the corpus
with pinned sensor hashes (the crop-mode framing is left to #754), bumps
RENDER_CACHE_VERSION to 43, and updates cr3.md, raw-parity (235 verified
models), coverage, gaps, the checklist rows, ROADMAP, target-app-parity and
What's New.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A Sony body shot in Adobe RGB writes ColorSpace Uncalibrated with InteropIndex R03 in the raw's EXIF and leaves the
embedded JPEG untagged, so the look was fitted against the JPEG decoded as sRGB. The EXIF fallback now applies to
all makes (the Nikon maker note still wins); DNGs are skipped, since their previews come from the converter.
Canon EOS R50 C-RAW files reach QP 168 (quant 44, step 102, measured about 100). The quantization step and the adaptive QP map range accept it; anything above stays rejected. Bumps RENDER_CACHE_VERSION since such files change from the embedded preview to a raw decode.
The 2668-entry table read off the black-box oracle's output is gone.
arw6_curve.rs now builds the 4096-entry table at first use from our own
measurement: piecewise linear through knots u_k = 128 k, c_k = u_k for
k <= 11 and 356.89 + 655.646 ln(k - 6) above, rounded half up, fitted to
the 7 same-exposure lossless (LJ92) / ARW6 pairs of the CC0 ILCE-7RM6 set.
A unit test pins the whole table (sum 27776336, weighted 92792980288),
its anchors, the identity through 1408 and strict monotonicity.
The corpus sums become regression values of the new output; the 12-bit
codes before the curve stay verified against the oracle (Phase 0).
docs/arw6-compression.md describes the curve, how it was measured, its
uncertainty, the structure and x2-scale evidence, why it is not in the
file, and the after-the-fact comparison with the oracle's curve; the
tracker and docs no longer call the output bit-exact against LibRaw.
Co-Authored-By: Claude <noreply@anthropic.com>
docs/superpowers/ held the working spec and plan, which pointed at the
local plan/ folder. The ARW6 module docs now cite the sections of
docs/arw6-compression.md, which gains the facts they relied on from the
spec: the NP field of word 0, the stream header word layout and how the
first half's residual lines are counted.
Co-Authored-By: Claude <noreply@anthropic.com>