PureJsImage v0.17.0

Added

• Added an explicit optional Rust/WASM WebP accelerator with scalar and SIMD modules for VP8 row
  color conversion, VP8L predictor and color transforms, lossless encode transforms, and lossy
  RGBA or gray to YUV420 conversion. Measured RGB input keeps the existing paired-row TypeScript
  conversion because it is faster end to end. The TypeScript codec still owns RIFF parsing, entropy
  coding, limits, metadata, and bounded row orchestration. Module or kernel failures fall back to
  the TypeScript path. The pull-request corpus workflow now forces scalar and SIMD WebP runs and
  checks exact decode pixels, deterministic encode bytes, and actual kernel use.
• Added a pull-request-only Imazen corpus workflow for JPEG, PNG, WebP, TIFF, GIF, and BMP. The
  workflow checks out the corpus at a pinned commit, runs every file in an isolated process, fails
  on per-file behavior changes, and uploads the generated reports for review.
• Added forced scalar and SIMD Imazen lanes for JPEG, PNG, and WebP. Eligible inputs must execute
  the selected WASM kernels. JPEG and PNG decode pixels remain exact, deterministic scalar and PNG
  encoders require byte parity, and the SIMD JPEG AAN encoder keeps its existing PSNR and size gate.

Changed

• Refreshed the stable-codec, cross-library web-codec, JPEG, PNG, and WebP WASM benchmarks, package
  metrics, result index, README charts, website charts, and accelerator benchmark pages after the
  current codec and runtime work.
• Node orientation and rotation now use lazy chunked memory by default. Pass
  { temporaryFiles: true } as the second argument to createImageLibrary() to opt into a lower
  process-RSS file spool. The opt-in path probes file creation, writing, reading, and truncation
  before consuming image rows. Failed setup or later file writes fall back to memory. The previous
  64 MiB Node memory fallback ceiling has been removed; normal image limits still apply.
• The optional WebP WASM accelerator now uses four-pixel SIMD for VP8L color transforms, fuses the
  common inverse color, predictor, and subtract-green row sequence into one call, and specializes
  uniform mode-11 prediction. Seven paired 4000x3000 lossless WebP trials reduced the median from
  617.90 ms to 518.50 ms, with exact output and a 2.00% peak RSS reduction.
• The optional Rust/WASM JPEG encoder now uses a separable f32 AAN transform, precomputed
  reciprocal quantization factors, and four-pixel SIMD RGB and RGBA compositing and color
  conversion. Unused SIMD scratch arrays were removed. Seven paired end-to-end trials reduced the
  png-to-jpeg workflow median by 24.58% with the output quality, size, and correctness gates
  unchanged.
• The optional Rust/WASM JPEG decoder now skips the full IDCT for DC-only blocks. Seven paired
  end-to-end trials reduced the jpeg-to-png workflow median by 5.64% with exact output
  correctness and a 0.46% peak RSS reduction.
• JPEG WASM exactness now covers DC-only inverse-transform half-integer boundaries and scalar
  encoder values immediately below a positive half-integer. The Rust decoder now validates the
  end-of-scan marker after its final MCU while still allowing legal bytes after EOI.
• Rust/WASM JPEG and PNG builds now reserve a bounded 128 KiB stack, trap on unexpected panics,
  and reject out-of-bounds or overlapping JPEG buffers before reading or writing memory. Initial
  accelerator memory remains within four WebAssembly pages.
• The Imazen corpus command now accepts --baseline <directory> and exits with an error when a
  file changes outcome, completed stage, structured error, diagnostic, child exit code, or signal.
  Timing and machine metadata remain outside the comparison.

PureJsImage v0.16.0

Added

• Added the purejsimage/geo raster API with CRS and affine grid metadata, resolution levels,
  bands and non-spatial axes, bounded pixel and world views, target-grid comparison, caller-supplied
  coordinate transforms, inverse-mapped reprojection, and the existing band math, terrain,
  statistics, histogram, and line-profile operations.
• Added lazy geo readers for GeoTIFF and COG, GeoZarr v2 and v3, world-file TIFF, JPEG, and PNG,
  georeferenced ENVI, Esri ASCII Grid, SRTM HGT, and classic NetCDF CDF-1 and CDF-2 files with
  regular rectilinear CF coordinates. The readers preserve native sample values, selected
  multidimensional axes, cancellation, source ownership, and bounded region access where the
  source format permits it.
• Added a generated geo compatibility manifest and deterministic CI benchmarks for remote COG,
  sharded and multidimensional GeoZarr, ENVI, classic NetCDF, and identity reprojection. The
  benchmark records request, byte, cache, chunk, shard, pixel, timing, and memory evidence while
  enforcing selective reads and output correctness.
• Added the /geo/ browser demo with deterministic local fixtures, public Kentucky aerial imagery,
  a USGS Grand Canyon elevation COG, pan and zoom controls, hillshade, live range telemetry, clear
  loading and error states, and copyable examples for the active GeoTIFF or GeoZarr view.

Changed

• The installed-package Geo consumer now runs deterministic range-backed GeoTIFF and
  multidimensional GeoZarr workflows through the public Geo, scientific, and tile-runtime entries.
  It also checks stable resource identity, cancellation ownership, exact cleanup, reprojection, and
  a browser Worker bundle. The Geo architecture guide now documents this application pattern.
• GeoTIFF overview levels now derive their affine transform when an overview repeats CRS metadata
  without its own model transform. Hillshade reads its required one-pixel source border without
  weakening the 196,608-pixel output limit, and geo views preserve exact nodata values and source
  level identity. Geo reprojection policies, serialized plans, and result provenance now preserve
  caller-supplied full-range int64 and uint64 nodata as canonical decimal strings.

PureJsImage v0.14.0

PureJsImage 0.14.0 adds deterministic, bounded numeric-raster analysis plans and tile executors for band math, normalized difference, linear combinations, subtraction, terrain slope/aspect/hillshade, regional statistics and histograms, line profiles, explicit target-grid resampling, and caller-supplied inverse reprojection.

Plans record raw/scaled value handling, units, nodata, algorithm versions, coordinate-transform identity and accuracy, and explicit tile and memory limits without adding runtime dependencies or full-raster materialization.

The browser convert demo also keeps original and converted images side by side on large screens and includes constrained AVIF output.

PureJsImage v0.12.0

PureJsImage 0.12.0

Added

• Added a deterministic Cloud Optimized GeoTIFF compatibility corpus, checked compression matrix,
  public bounded inspectCog() structural report, explicit compression ID/name errors, and a
  simulated-HTTP-range viewport benchmark that asserts overview selection and reports requests,
  bytes fetched, cache hits, first decoded tile-block latency, and decoded pixels.

• Added typed GeoTIFF spatial references to ordinary TIFF scientific dataset descriptors and
  SubIFD levels, including CRS authority/code and citation, six-parameter pixel-to-model affine,
  invertible inverse, model bounds, pixel-is-area/point semantics, scalar or component GDAL nodata,
  and JSON-safe source metadata. Region reads remain in raster pixel coordinates, and malformed
  optional georeferencing does not block native TIFF pixels.

• Expanded the official common-web (web-codecs) suite with JPEG→AVIF, JPEG→lossy WebP, AVIF
  crop+resize, the Lambda JPEG thumbnail, progressive JPEG resize, and lossless WebP alpha decode.
  The published speed and memory SVG charts include those jobs, and the README now shows both
  charts.

• Added the explicit purejsimage/scientific/readers/ome-zarr reader for OME-NGFF 0.4 and 0.5
  image multiscales on first-party Zarr v2 and v3 directory stores. Selected planes fetch only
  intersecting regular or sharding_indexed chunks through the existing companion resolver; bytes,
  gzip, zlib, zstd, crc32c, transpose, shuffle, and Blosc 1 (LZ4/zlib/zstd/memcpy) are implemented;
  missing chunks become fill values. Sibling and root NGFF labels become separate datasets with
  image-label colors; plate wells become one dataset per field. A ZIP archive with root-level
  zarr.json or .zgroup opens as a single-file store without companions. Partial last chunks,
  sharded inner endian, Blosc split streams, and the C-Blosc compressor enumeration are handled.
  A pinned IDR 6001240 coarsest-plane slice cross-checks NGFF 0.4 Blosc/LZ4 against 0.5 sharded
  Blosc/zstd. The reader also accepts omitted v3 chunk_key_encoding, exact integer and hex
  fill_values, F-order padded last chunks, numeric OMERO colors, UTF-8 BOM metadata, and
  big-endian shard indexes, and rejects overlapping shard payloads. any-axis-pair planes are
  packed into destination row order. Zarr v2 fill_value: null decodes present chunks and fails
  on absent ones instead of inventing zeros; a present zero-byte chunk is malformed; missing
  shard inners use only the uint64 all-ones sentinel pair. Directory dataset identity includes
  defining metadata plus a session store id; ZIP identity stays the archive source. Chunk lookup
  uses a bounded LRU so maxOpenSources is not a lifetime visit limit. Clipped last chunks no
  longer fail when the nominal chunk exceeds the decode budget; empty optional .zattrs, trailing
  dataset slashes, bare one-byte NumPy dtypes, numcodecs.* codec ids, and case-insensitive
  NaN/Infinity fills are accepted. bioformats2raw 0/, 1/, … series roots and a single nested
  ZIP store prefix are supported; ambiguous multi-root ZIPs stay rejected. Series discovery also
  works without a root .zattrs file, bioformats2raw.layout may be a numeric string, and
  seriesCount counts series groups rather than every collected dataset. ZIP stores named
  *.zarr / *.ome.zarr probe as OME-Zarr, macOS __MACOSX/ sidecars no longer make a unique
  nested root look ambiguous, a root labels list does not hide bioformats2raw series, and extra
  integer series beyond maxDatasets raise LIMIT_EXCEEDED. ZIP probes use name-plus-magic
  evidence and do not open the archive. Generic Zarr v2 groups are not detected without NGFF
  attributes. Zarr storage fill is recorded as zarrFill rather than noDataValue; v3 rejects
  null and non-numeric integer fills; shard-index codecs must declare endian; nonempty
  storage_transformers are unsupported; NGFF axes must be time, then channel or custom, then
  2–3 spatial axes; OMERO colors must be six hex digits or 0–0xffffff; and the chunk cache is
  bounded by maxCachedChunkBytes. Pinned IDR slices now include 6001240 sibling labels, an
  IDR0010 0.5 plate well, an IDR0001 0.4 plate field, and an IDR0101 translation image. BloscLZ,
  Snappy, bitshuffle, tables, RFC-9 zip-comment/jsonFirst requirements, nonempty storage
  transformers, repeated transpose codecs, and writers remain explicit unsupported operations.
  ZIP members are limited before decompression; sharded plane reads resolve each shard once;
  bioformats2raw.layout must be 3; and directory calibration cites the resolved metadata
  resource id. Plane-session caches stay bounded and layer over the persistent chunk LRU;
  a source larger than maxCachedChunkBytes is held at most as one transient shard.

Changed

• Speed up first-party AVIF decode after the retained inverse-transform skip: decode equiprobable
  bits without a throwaway CDF, skip zero-coefficient dequant, upsample interior 4:2:0 chroma
  without edge clips, renormalize the arithmetic coder with integer shifts, clamp YUV bytes with
  integer rounding, hoist SGR prefix-row bases, and inline the common 8-bit 4:2:0 convert. Local
  avif-fox-resize-jpeg fell from 589 ms to 500 ms (−15%) with an exact output hash; neighbor
  avif-fox-full-png improved 11%. The Imazen AVIF survey stayed 36/36 decoded with maximum RGB
  error 2; post-filter, color, high-bit, tile, q-matrix, and common-photo oracles were unchanged.
• Speed up first-party AVIF inverse transforms by reusing residual scratch and skipping 1D
  transforms on all-zero rows and columns. Local avif-fox-resize-jpeg fell from 621 ms to 589 ms
  (−5%) with an exact output hash. The Imazen AVIF survey stayed 36/36 decoded with maximum RGB
  error 2; post-filter, high-bit, tile, q-matrix, and common-photo oracles were unchanged.
• Speed up first-party AVIF loop restoration by specializing the 8-bit SGR box filter, restoring
  8-bit SGR/Wiener as unit-width tiles in 8-row bands, and applying interior Wiener 7-tap filters
  directly from the CDEF plane. Local avif-fox-resize-jpeg fell from 838 ms to 623 ms (−25%) with
  an exact output hash; neighbor avif-fox-full-png improved 18%. High-bit restoration stays on
  4-wide 4-row tiles so Int32 prefix squares do not overflow. The Imazen AVIF survey stayed 36/36
  decoded with maximum RGB error 2.
• Fuse 8-bit AC Huffman prefix decoding into scaled JPEG block decode. Local jpeg-resize-1200
  fell from 543 ms to 511 ms (−7%) with an exact bitstream hash; neighbor jpeg-crop-resize
  improved 3.9% and northstar-photo-pipeline improved 3.4%. Imazen JPEG stayed 39 pass / 2
  unsupported / 167 rejected-safely / 46 accepted.
• Speed up JPEG resize-to-encode pipelines by unrolling the scale-2 4×4 IDCT, skipping vertical
  chroma bilinear on 4:2:2 (chroma already has full vertical resolution), and unrolling the encoder
  DCT. Local jpeg-resize-1200 fell from 715 ms to 543 ms (−24%) with an exact bitstream hash;
  neighbor jpeg-crop-resize improved 18%. Imazen JPEG stayed 39 pass / 2 unsupported / 167
  rejected-safely / 46 accepted.
• Speed up scaled JPEG decode by skipping unused AC store after the last zigzag the reduced IDCT
  reads, fusing leftover Huffman bit skips, unrolling the scale-4 2×2 IDCT, stopping restart
  indexing once the crop target is passed, and inlining leftover-AC skip on the entropy reader.
  Official northstar-photo-pipeline fell from 1181 ms to 802 ms (−32%) on top of the previous
  scaled-IDCT work, with an exact bitstream hash. Neighbor jpeg-crop-resize improved 13.5%;
  jpeg-resize-1200 kept its exact hash. Imazen JPEG stayed 39 pass / 2 unsupported / 167
  rejected-safely / 46 accepted.
• Speed up large JPEG crop-then-resize pipelines by snapping unaligned decoder crops to a
  containing scale-aligned box so scaled IDCT can run. Official northstar-photo-pipeline fell
  from 2857 ms to 1181 ms (−59%). Pixel samples stayed inside the documented ±8 tolerance; the
  lossy JPEG bitstream changed as expected. Aligned neighbors kept exact hashes. Imazen JPEG
  stayed 39 pass / 2 unsupported / 167 rejected-safely / 46 accepted.
• Speed up JPEG-to-PNG conversion by recycling decoder row buffers after encode, specializing
  adaptive PNG filter scoring for RGB8, and skipping bilinear luma interpolation when luma is
  already full resolution. Official jpeg-to-png fell from 552 ms to 444 ms (−20%) with a 6.5%
  peak-RSS drop on the recycle path; neighbor jpeg-resize-1200 improved 4.5%. Output hashes and
  the Imazen JPEG/PNG corpora were unchanged.
• Fail full official web-codecs and competitors benchmark runs when any matched pass/pass wall
  median is more than 10% slower than the published headline snapshot. Single-workflow hillclimb
  runs are not gated.
• Publish web-codec and competitor benchmark charts as SVG instead of rasterizing them to PNG.
• Speed up first-party AVIF decode after the retained restoration gather/prefix work: hoist SGR
  prefixes, unroll Wiener and SGR blend, copy interior CDEF windows, reuse inverse-transform
  scratch buffers, specialize 4:2:0 chroma upsample, and specialize 8-bit Wiener rounding. Official
  avif-fox-resize-jpeg fell from 1452 ms to 803 ms (−45%) and avif-fox-full-png from 1573 ms to
  987 ms (−37%) on the 2026-08-17 web-codec snapshot, with unchanged output hashes.
• Speed up first-party lossless WebP encoding by replacing the shifting LZ77 candidate table with a
  ring buffer, recording match tokens once for histogram and bitstream emission, and scoring color
  cache sizes in a single scan. Isolated 1200x480 effort-4 encode of transparent-logo-1200x480
  fell from 166 ms to 100 ms with an unchanged bitstream.
• Speed up first-party lossy WebP encoding by adding the DC predictor during reconstruction instead
  of prefilling each 4x4, and by finalizing RGB8 chroma in the 2x2 write. Isolated 1200x900
  quality-80 enc...

PureJsImage v0.11.0

PureJsImage 0.11.0

Added

• Added the explicit purejsimage/codecs/web entry with the frozen allWebCodecs JPEG, PNG, WebP,
  and AVIF aggregate, plus a correctness-gated common-web benchmark profile with AVIF metadata,
  full-decode, resize, conversion, and external Sharp and jSquash comparison rows. The historical
  ordinary competitor profile remains unchanged.

• Completed the initial Milestone H interchange and detector set with explicit portable readers
  for RPL/RAW, EMSA/MAS, NRRD, MetaImage MHD/MHA, NIfTI-1/2 including bounded .nii.gz, NPY,
  NanoMegas BLO, processed Merlin MIB, and rectangular ANG/CTF orientation maps. The readers
  preserve native numeric samples and calibration evidence, provide bounded selected reads where
  their storage permits, reject recognized unsupported variants, ship as individual package
  entries, and pass generated structural, hostile-limit, packed-consumer, size, and real-Chromium
  coverage. Pinned RosettaSciIO files independently exercise RPL/RAW, ISO EMSA, BLO, and processed
  Merlin MIB without redistributing their GPL binaries. EDAX and Bruker BCF remain the roadmap's
  explicit later items.

• Added the public purejsimage/scientific/readers/ncem-emd reader for the fixture-proven
  Berkeley/openNCEM 0.2 subset. It accepts integer or decimal-string versions, discovers numeric
  groups below /data or /signals, preserves exact labeled coordinates and bounded scalar or
  array acquisition metadata, and performs selected HDF5 hyperslab reads. Three pinned real
  RosettaSciIO application files join generated hostile and independent h5py coverage without
  redistributing their GPL binaries; Direct Electron .de5 remains explicitly unsupported.

• Added the separate public purejsimage/scientific/readers/velox-emd E2 image reader with
  hierarchy-based probing, bounded per-frame JSON under an aggregate budget, separate detector
  datasets and frame axes, native scalar and complex data, preserved positive-half uncentered FFT
  storage, and explicit
  pruned-spectrum-image errors. Every metadata column is preserved, while calibration-critical
  conflicts fall back to index axes with a structured warning. Generated hostile fixtures and
  pinned TEM stack, DPC, and FFT files verify the reader without committing the GPL fixture
  binaries. Sparse Velox spectra remain gated on the Lab Viewer spectrum surface and explicit
  event-decoding contracts.

• Added structured calibration provenance to scientific axes, including validated embedded,
  sidecar, derived, and format-default evidence. GSF, FITS, MRC/CCP4, OME-TIFF, and Aperio SVS
  readers now identify the exact source resource and stable metadata locator behind each available
  physical axis.

• Added the public createImageCodecScientificReader() fallback adapter and individually
  importable PNG, JPEG, WebP, BMP, and JP2 scientific readers. They preserve codec pixels and
  component semantics, source identity, cancellation, block release ownership, and explicit reader
  registration without linking codecs into the base scientific entry.

• Added purejsimage/scientific/readers/tiff, a native-precision ordinary TIFF reader with labeled
  page axes, incompatible-series separation, SubIFD resolution levels, bounded normalized optional
  metadata, native N-channel layouts, and lower probe priority than OME-TIFF and Aperio SVS.

• Added bounded FEI SFEG/Helios and Zeiss SEM TIFF calibration profiles with fixture-backed physical
  axes, embedded evidence, normalized acquisition metadata, and non-fatal malformed-tag handling.

• Added a package-private bounded DM3/DM4 tag-tree indexer with checked 32-bit and 64-bit structure,
  separate payload byte order, recursive value descriptors, deterministic duplicate-name paths,
  lazy image payload spans, and size-limited metadata projection for the upcoming scientific
  reader.

• Added the public purejsimage/scientific/readers/digital-micrograph reader for supported DM3/DM4
  rank-2 through rank-4 scalar images and volumes, all signed/unsigned 8/16/32-bit and
  float32/float64 samples, fixture-proven packed BGRA, calibrated axes and intensity units,
  bounded namespaced Gatan metadata, and direct selected-region reads. Rank-1, complex,
  undocumented packed, encrypted, and external image content remains explicitly unsupported.

• Added a generated scientific-reader capability section covering every public reader descriptor,
  package export, hint, resource model, dataset kind, direct-range boundary, evidence, and fixture
  source. Added a pinned, checksum-verified RosettaSciIO DigitalMicrograph compatibility corpus
  workflow without redistributing its GPL-licensed fixture binaries in the package repository.

• Added evidence-gated DigitalMicrograph multidimensional semantics: ordinary X/Y/Z volumes,
  Gatan-tagged X/Y/energy EELS spectrum images, and C-ordered 4D-STEM exposed logically as
  scanX/scanY/kx/ky with direct kx/ky diffraction-plane reads. Ambiguous rank-4 arrays remain
  neutral. A pinned small DM4 volume and bounded HTTP-range verification of LiberTEM's 1.19 GB
  CC-BY-4.0 Zenodo 4D-STEM fixture cover the new mappings without committing the large file.

• Added honest one-axis scientific datasets with explicit native series-read capabilities, bounded
  ScientificSeriesBlock output, strict series request normalization, and a bounded row/column
  adapter over existing plane readers. Existing plane-only descriptors and readers remain
  unchanged.

• Added the public purejsimage/scientific/readers/tia-ser reader for FEI/Thermo TIA SER v528 and
  v544 scalar spectra, spectrum images, and image series. It indexes calibrated elements and valid
  counts without eager payload reads, preserves incomplete elements as bounded document metadata,
  provides direct canonical sample reads, and is verified against pinned real RosettaSciIO files
  without redistributing their GPL-licensed binaries.

• Added the public purejsimage/scientific/readers/tia-emi companion reader for portable embedded
  ObjectInfo XML, numbered SER resource composition, acquisition metadata, stable identities,
  and evidence-preserving reciprocal-space reinterpretation. Contradictory SER coordinates remain
  authoritative and are reported rather than overwritten.

• Added a package-private HDF5 file and address layer with legal user-block signature discovery,
  superblock versions 0 through 3, 2/4/8/16-byte integers, relocation-aware bigint addresses,
  lookup3 superblock checksum verification, and a bounded source-identity-aware metadata page
  cache. Legacy family, multi-file, and unknown drivers plus modern superblock extensions remain
  explicit unsupported boundaries while the object graph is implemented separately.

• Added the first package-private HDF5 object-graph slice with version 1 and 2 object headers,
  checksummed continuation chunks, compact hard and soft links, link-info storage descriptors,
  bounded legacy symbol-table groups with local heaps and B-tree v1, and configurable
  hostile-metadata limits. A revision- and SHA-256-pinned HDF Group fixture verifies the real legacy
  group path without adding its binary to the repository.

• Continued the package-private HDF5 object graph with checksummed fractal-heap headers, root direct
  and recursive indirect managed blocks, seven-byte managed heap IDs, and type-5 B-tree v2 leaf and
  internal nodes for modern dense groups. Generated hostile fixtures cover hard, soft, ASCII, UTF-8,
  creation-order, bounds, cycles, ordering, and corruption behavior. A second pinned HDF Group file
  verifies a real 40-record dense index and the explicit external-link boundary. Huge and tiny heap
  objects, filtered heaps, and the secondary creation-order index remain explicit deferred
  variants; bounded graph traversal, attributes, dataset layouts, chunk indexes, required filters,
  and exact selected reads are implemented by the later package-private D2-D6 layers. General
  public HDF5 reader claims remain intentionally absent.

• Completed the fixture-proven package-private HDF5 D3-D6 substrate: bounded datatypes,
  dataspaces, attributes, compact/contiguous/chunked layouts, every required classic and modern
  chunk index, raw/Deflate/Shuffle/Fletcher32 filters, exact fill handling, local/HTTP parity, and
  cancellable selected dataset blocks. Contiguous rank-2 strided selections now batch many rows
  under independent input-span, output, and read-operation caps.

• Added public Nanonis SXM, Igor Binary Wave v5, Digital Surf SUR/PRO, and ISO 5436-2 X3P surface
  readers plus a bounded ZIP/ZIP64 container. Committed test-only surface binaries now have a
  machine-readable per-file manifest with exact source revision/path/URL, license, attribution,
  redistribution rationale, SHA-256, and oracle.

Changed

• Documentation pages keep wide tables, code samples, comparison matrices, and chip navigation
  inside named mobile scroll regions. The homepage comparison now leads with a compact six-dimension
  summary on narrow screens and keeps the full TIFF matrix in an expandable, prerendered section.
• The README now leads with measured whole-slide, scientific explorer, and web codec memory visuals,
  and the generated scientific reader block is a family summary that links to the format reference.
• Hardened scientific probing and metadata semantics after review: HDF5 dialect probes use exact
  eight-byte signature reads and pass the verified user-block offset into dialect opening; shared
  HDF5 loads isolate caller cancellation, preserve late-waiter coalescing, coalesce graph accounting,
  and cannot repopulate after close; Velox enforces its JSON budget across the complete document;
  NIfTI ignores scl_inter when scl_slope is zero, classifies affine columns relative to their
  own scale, and rejects non-finite header fields explicitly; and DM, TIA SER, Velox, and NIfTI
  omit physical units/evidence when numeric calibration is incomplete while re...

PureJsImage v0.9.0

Added

• Expanded TIFF decoding with unsigned 6-, 10-, 12-, and 14-bit grayscale, 2-, 4-, 10-, 12-,
  and 14-bit chunky or planar RGB, native packed horizontal prediction, direct 16-bit output,
  and five-sample CMYK plus associated or unassociated alpha while retaining bounded segment
  memory.
• Added explicit WebP-in-TIFF composition through
  createTiffCodec({ embeddedCodecs: [webpCodec] }); the default TIFF codec and root package do
  not import or activate WebP automatically.
• Added a complete development-only TIFF dependency matrix plus isolated raw full-image and region
  benchmark operations with source-read, maximum decoded-block, external-memory, and ArrayBuffer
  measurements.
• Added a generated evidence-backed TIFF library comparison across the README and documentation,
  backed by pinned versions and a 154-file isolated-process conformance run that reports exact
  pixels, mismatches, unsupported cases, errors, timeouts, crashes, native scientific rasters, and
  malformed-input behavior separately.
• Expanded the browser demo from 8 to 38 public examples spanning ECCI scanning electron
  microscopy, OME-TIFF microscopy dimensions, FLIM, SPIM, high-content screening, pathology,
  scientific JPEG 2000, high-bit-depth AVIF, and additional unusual codec layouts.
• Added a published llms.txt with a capability-manifest-generated codec map, complete quick API
  reference, runtime and safety boundaries, and migration guidance for Jimp, Sharp, image-js,
  jSquash, GeoTIFF.js, and UTIF.js; every website footer now links the guide and sitemap.
• Added TIFF signed 8-/16-bit and IEEE float16/float32/float64 grayscale and RGB decoding with
  native-precision raw pixel blocks, deterministic display ranges, bounded Predictor 2 and
  floating Predictor 3 reversal, and browser-compatible display conversion.
• Added unsigned 24-/32-/64-bit TIFF grayscale and RGB decoding with canonical gray32,
  rgb32, gray64, and rgb64 raw blocks, exact values above JavaScript's safe-integer range,
  bounded wide Predictor 2 reversal, and deterministic display conversion without an 8-/16-bit
  raw intermediate or per-pixel BigInt arithmetic.
• Added bounded TIFF SGILog and SGILog24 decompression with native CIE Y/XYZ float32 pixel
  blocks, logarithmic luminance and chroma reconstruction, deterministic CCIR 709 gamma-2
  display conversion, exact segment validation, and browser-compatible strip and tile decoding.
• Added explicit TIFF top-level frame and reduced-resolution SubIFD selection for Classic TIFF
  and BigTIFF, with selected-level metadata, bounded IFD graph traversal, cycle rejection,
  alias-safe shared-directory reuse, global directory limits, and no reads from unselected pixel
  segments.
• Added 16-bit TIFF palette decode with exact full-range ColorMap scaling, plus signed 8-/16-bit
  and float16/float32/float64 CMYK display conversion using declared sample ranges or deterministic
  full-type defaults, bounded segment output, and independent pixel validation.
• Added 8-bit TIFF CIELab conversion from D65-referenced Lab* to sRGB, bounded CMYK
  lut16 A2B0 ICC-profile conversion with profile precedence over numeric display, and
  FillOrder=2 normalization for non-fax strips and tiles before predictor reversal without
  mutating aliased source buffers.
• Added a public TIFF document layer at purejsimage/tiff with stable top-level/SubIFD graphs,
  absolute IFD offsets and lookup, bounded defensive-copy private metadata reads, cached immutable
  typed tags with per-call limits, per-directory display and native raster decoders, deterministic
  third-party profile registration with isolated detector failures and ambiguity rejection, and
  typed explicit TiffProfile<T> opening.
• Added native-precision planar or interleaved N-channel RasterBlock output plus explicit
  rasterToPixels() range mapping, so scientific samples no longer require an implicit RGB
  interpretation.
• Added purejsimage/scientific OME-TIFF datasets with validated Z/C/T dimensions, channel and
  physical-pixel metadata, explicit or implicit TiffData plane mappings, separate-channel
  assembly, and reduced-resolution SubIFD selection.
• Added a generic WholeSlideImage contract and first-party Aperio SVS profile with bounded
  pyramid/associated-image region reads, MPP and objective metadata, reusable JPEG 2000 codestream
  composition, and TIFF compression tags 33003/33005. The pinned irreversible Aperio tile is
  independently validated against OpenSlide within two 8-bit code values.
• Added a separately compiled Leica SCN single-area profile example that imports only published
  package entries, with an automated boundary check suitable for independent vendor packages.
• Added a dedicated TIFF documentation page covering the complete decode matrix, scientific and
  whole-slide APIs, third-party profiles, canonical output, memory model, and unsupported
  boundaries; reduced the README TIFF and Zstandard sections to direct documentation links and
  refreshed the measured bundle and installed-size tables.
• Added a reusable first-party Zstandard decompressor at purejsimage/compression/zstd with
  explicit output and window bounds, raw/RLE/compressed blocks, Huffman and FSE entropy decoding,
  repeated tables and offsets, frame checksums, and structured malformed-input failures.
• Added bounded TIFF Compression=50000 strip and tile decoding through the reusable Zstandard
  backend, preserving existing predictor, sample, and pixel processing.
• Replaced simple uncompressed TIFF output with a canonical Classic little-endian 8-bit RGB/RGBA
  encoder using independently Deflate-compressed, horizontally predicted strips, automatic
  roughly 128 KiB strip planning, strict extensible strategy options, bounded raw strip scratch,
  compatible RGB ICC preservation, and exact ImageMagick/LibTIFF interoperability validation.
• Added bounded first-party TIFF LERC2 and LERC-plus-Deflate segment decoding with mask handling,
  native numeric sample reconstruction, independently validated byte/float pixels, and structured
  rejection of corrupt headers, dimensions, masks, checksums, and codec metadata.
• Added first-party GeoTIFF model, coordinate, bounding-box, GeoKey, GDAL metadata, and nodata
  helpers over the public TIFF document API, with exact affine and tiepoint semantics.
• Added a bounded HttpRangeSource with request coalescing, validator-protected reads, byte-capped
  LRU caching, response cancellation, and selective COG-style region access that does not fetch
  unrelated tiles.
• Expanded structured TIFF output with explicit rows per strip, tiled RGB/RGBA segments, automatic
  or explicit BigTIFF selection, multi-page top-level IFD chains, and reduced-resolution SubIFD
  pyramids, independently reopened through GeoTIFF.js.
• Added a development-only TIFF conformance harness that scores PureJsImage, GeoTIFF.js, UTIF.js,
  image-js, and Jimp against independent sharp/ImageMagick RGBA output with isolated memory,
  timeout, crash, exact-pixel, and error reporting.

Changed

• Moved OME-TIFF and raster helpers, Aperio SVS, and GeoTIFF profiles out of the root and browser
  entries into purejsimage/scientific, purejsimage/pathology, and purejsimage/tiff; moved
  HttpRangeSource into purejsimage/sources/http-range so the core bundle does not retain
  specialized TIFF workflows or the optional HTTP adapter.
• Relabeled TIFF conformance as an unreleased commit-pinned workspace snapshot and made decoded
  coverage the compact comparison headline, with exact pixels and failure counts reported
  separately.

PureJsImage v0.8.0

Added

• Extended the isolated Imazen codec-corpus harness and independent baselines across JPEG, PNG,
  WebP, TIFF, GIF, and BMP, covering complete decode-to-PNG round trips, structured rejection,
  timeouts, crashes, memory failures, upstream categories, and format-specific feature groups.
• Added an explicitly imported first-party Rust/WASM PNG accelerator for common non-interlaced
  8-bit grayscale, RGB, and RGBA scanline decode and encode, retaining native runtime zlib,
  bounded-row memory, scalar fallback, and the TypeScript reference for every ineligible workload.
• Added a separate purejsimage-wasm competitor benchmark engine and seven-engine speed, memory,
  and quality reports so the explicitly registered JPEG/PNG accelerators can be compared directly
  with the unchanged default TypeScript package.
• Added an explicit optional first-party Rust/WASM baseline JPEG decoder that fuses entropy decode,
  IDCT, chroma upsampling, and RGB conversion while preserving bounded MCU-row output, exact
  TypeScript parity, lazy warm reuse, and clean fallback for unsupported or unhelpful workloads.
• Added optional first-party scalar and SIMD Rust/WASM baseline JPEG encoders with bounded MCU-row
  input, JavaScript-owned sink backpressure, RGB/RGBA/grayscale and 4:2:0/4:2:2/4:4:4 parity, explicit
  workload selection, plus an ABI-compatible SIMD JPEG decoder artifact with scalar fallback.
  Specialized entropy writes, validated-row input reads, chroma sampling, Huffman lookup, and paired
  decoder upsampling reduce pinned warm encoder time by 61.7%-66.0% versus TypeScript and pinned warm
  decoder time by 58.4%, while retaining the same output contracts and bounded memory.
• Added first-party refinement-based progressive JPEG encoding for grayscale and 4:2:0, 4:2:2,
  and 4:4:4 YCbCr output, with compact coefficient limits, per-scan restart markers, independent
  pixel validation, browser coverage, and isolated runtime/RSS/size benchmarks.
• Added jSquash as an isolated WebAssembly competitor using its pinned JPEG, PNG, WebP, and resize
  packages, with documented Node WASM initialization, explicit unsupported classifications,
  startup/package-footprint measurements, a codec-matched bundle comparison with exact package
  versions, validated workflow results, and regenerated charts.
• Added premultiplied-RGBA PSNR against independent exact-area references to quality-enabled
  competitor workflows, with quality recorded outside timing and peak-RSS sampling, included in
  Markdown and JSON reports, and published as a dedicated comparison chart.
• Added a fully client-side browser conversion demo with content-based format detection, optional
  transforms, honest timing and memory reporting, and an artifact-only GitHub Pages deployment
  that keeps its generated all-codec bundle out of repository history.

Changed

• Moved HEIF/HEIC decode to the explicit
  purejsimage/codecs/experimental/heic entry, removed it from allCodecs and
  the default browser demo, and documented that MIT grants no third-party HEVC
  patent rights.
• Documented when native Sharp is the better performance choice and added measured Lambda sizing
  guidance: 256 MiB completed every pinned 12-megapixel workflow, but 1024 MiB cut the JPEG-to-WebP
  warm operation from 10.6 seconds to 2.5 seconds while peak use remained about 120 MiB.

Fixed

• Kept the nested packed-declaration smoke test operational during npm pack --dry-run, so the
  documented release gate validates package contents instead of inheriting npm's outer dry-run mode.

• JPEG decoding now treats sampling factors as one for single-component non-interleaved scans,
  tolerantly preserves completed progressive coefficients when a partial scan reaches a DHT, SOS,
  or EOI boundary, and decodes AVI1/MJPEG baseline frames that omit standard Huffman tables. Strict
  progressive decoding continues to reject partial entropy.

• Recognized 12-bit and arithmetic-coded JPEG inputs now fail with UNSUPPORTED_OPERATION,
  separating deliberate codec boundaries from supported-subset failures in the Imazen conformance
  baseline.

• TIFF decoding now defers BigTIFF offset validation until an inline value actually needs an
  external offset, recognizes legacy LSB-packed LZW streams with late code-width changes, accepts
  legacy tile tables in strip tags, bounds padded final YCbCr LZW strips, reconstructs multi-strip
  old-style JPEGs with omitted legacy fields, and decodes one-dimensional Group 3 fax rows without
  EOL markers. Recognized 64-bit sample layouts remain structured unsupported boundaries. BMP RLE4
  now accepts the single encoded padding pixel used for odd-width scanlines while retaining overrun
  rejection.

• Animated GIF pixel decode now fails with UNSUPPORTED_OPERATION instead of silently discarding
  animation; callers can explicitly request the supported first image with open(input, { frame: 0 }).

• Baseline JPEG restart recovery now defaults to tolerant decoding for malformed real-world files;
  pass open(input, { tolerantDecoding: false }) to require strict restart sequencing. Explicit
  Rust/WASM decoding now implements the same bounded recovery instead of being skipped. JPEG and
  PNG decode accelerator setup or midstream failures now resume through their TypeScript decoders
  without duplicate rows. Focused scalar and SIMD PNG regressions keep trailing IEND data,
  full-range cICP, and ICC v4 RGB mAB color conversion at parity without silently falling back.
  The same corpus-driven work completed those compatibility fixes for the TypeScript reference.

• Removed ambient Buffer references from the Node entry's published declarations so strict
  TypeScript consumers can compile the zero-dependency package without installing @types/node.

• Changed the default resize kernel from bilinear to scale-aware Lanczos 3 so ordinary downscales
  no longer discard most source samples and alias heavily; strong downscales now use bounded
  streaming box pre-shrink before the final Lanczos pass, format-specialized horizontal kernels,
  opaque-RGBA handling, and a fixed retained-row ring instead of a map. Bilinear remains available
  as an explicit faster, lower-quality option, and benchmark reports identify cross-kernel
  comparisons. On the pinned profiles, the 4× PNG resize fell from 1,253 ms to 530 ms and the
  10× 100-megapixel downscale fell from 8,455 ms to 2,408 ms.

• Allowed the browser demo to convert the supported primary image from iPhone-style MPF JPEGs while
  warning that auxiliary images and gain maps are not preserved; true animated inputs remain blocked.

• Reduced progressive JPEG output with two-pass, scan-specific optimized Huffman tables; the pinned
  quality-80 benchmark moved from 3.29% larger than baseline to 7.26% smaller with identical decoded
  PSNR.

• Corrected AV1 quantization-matrix coefficient-axis lookup and adjusted matrix
  dimensions for 64-point transforms, with q30-q90 YUV and displayed-RGB gates
  against independent libaom, dav1d, Sharp, and Chromium decoders.

• Reduced lossless WebP output for screenshot-style pixels with block-adaptive VP8L predictors,
  adaptive color-cache codes, and a deeper bounded match search, while retaining exact independent
  libwebp pixel validation and documenting the fixed match-table memory cost.

• Fixed progressive JPEG AC refinement after a zero-run-length symbol so common Sharp/libjpeg
  output decodes correctly and can be resized and converted to WebP in Node.js and browsers.

PureJsImage v0.7.0

Added

• Completed the common static 8-bit Huffman JPEG reference before optional WASM work: incremental
  entropy input, restart-aware region seeking, chroma-aware bilinear upsampling, SOF1 and sequential
  multi-scan decode, 4:4:0 and 4:1:1 fixtures, native grayscale and restart-marker output, hostile
  input limits, and provider-neutral parity vectors.
• Added native 1/2, 1/4, and 1/8 JPEG IDCT output, decoder-driven scale selection for safe
  full-frame downscales, baseline/progressive/restart/subsampling regressions, and an isolated
  runtime, RSS, avoided-pixel, and output-error benchmark against the full-resolution path.
• Added an explicit modern-browser entry with File/Blob, ArrayBuffer, Uint8Array, Blob output,
  custom sinks, browser-safe TypeScript declarations, and a browser bundle gate that rejects Node
  built-ins across the full codec graph.
• Added browser PNG encoding through CompressionStream and bounded browser rotation/orientation
  storage through OPFS, a lazy 64 MiB chunked-memory fallback, and IndexedDB for larger transforms.
• Added HEIF/HEIC EXIF and RGB ICC preservation for conversion workflows, including orientation
  normalization, plus structured ICC and nclx color-profile metadata for HEIF and AVIF.
• Refreshed the static GitHub Pages documentation with responsive guides, API and codec references,
  benchmark tables, memory-model explanations, and clearer capability boundaries.
• Documented the progressive JavaScript, optional WASM, and future WebGPU backend roadmap,
  including capability probing, workload-based selection, bounded GPU memory rules, runtime
  fallback policy, correctness gates, and a staged browser-first delivery plan.
• Added a reproducible bundle and deployment-footprint comparison for PureJsImage, Jimp, image-js,
  and Sharp, with a matched common codec set and explicit accounting for Sharp's native payload.
• Added one machine-readable codec capability manifest that generates README, detailed codec,
  website, public JSON, and compatibility-test expectations, with a stale-output CI gate and a
  repository agent skill for rolling capabilities out consistently.

Changed

• Made bundle and installed-package size reporting part of the mandatory npm run check gate.
• Reworked the root README into a shorter human-facing overview with a practical all-codec example,
  bundle sizes, codec support and compatibility links, benchmark charts, and direct routes to the
  API, benchmark, browser, temporary-storage, specification, and contributor details.
• Split Node file, Buffer, zlib, and temporary-file services from the portable codec and pipeline
  core while preserving the existing Node API and measured transform performance.
• Made Node.js and modern-browser portability an explicit project goal and release requirement.
• Reused a single Node file descriptor across detection, metadata, and execution reads, and avoided
  redundant copies when stable sources fill codec regions.
• Surfaced the measured 191.8-211.0 MiB JPEG 2000 decode peak and its 512 MiB Lambda-tier guidance
  in the README and codec documentation.

Fixed

• Reduced first-party lossless WebP output for graphics by adding predictor and subtract-green
  transforms, LZ77 references, and adaptive Huffman trees, with exact Sharp/libwebp pixel checks
  and an isolated size and peak-memory benchmark for the documented full-frame encoder state.
• Decoded the quantization matrices and block delta-Q syntax used by default Sharp/libaom AVIF
  output in the restricted opaque 8-bit YUV 4:2:0 path, with exact q50 and q80 YUV agreement
  against both dav1d and libaom while delta loop-filter syntax remains explicitly unsupported.
• Restricted SVG recognition to the document's first element so HTML with inline SVG and JSON text
  mentioning <svg are no longer misreported as SVG input.

PureJsImage v0.6.0

Added

• Added first-party JPEG 2000 / JP2 decoding for common Part 1 grayscale and RGB still images,
  including reversible 5/3 and irreversible 9/7 reconstruction, all five progression orders,
  multiple tiles, 1-16-bit unsigned samples, strict container validation, public pipeline
  conversion, and checksum-pinned fixtures from independent encoders.
• Added public-domain real-photograph JP2 fixtures with exact OpenJPEG oracle hashes, structured
  container and codestream corruption regressions, a deterministic JP2 mutation campaign, decoded
  allocation-limit tests, and isolated cold/warm RSS gates for the documented full-frame fallback.
• Added ordered, composable spatial stages: crop-after-resize, multiple resize stages, clockwise
  rotate() with arbitrary-angle bilinear sampling, vertical flip(), and horizontal flop().
• Added opt-in keepExif() and keepIcc() pipeline operations with JPEG, PNG, and WebP metadata
  round trips, TIFF ICC round trips, orientation normalization after pixel reorientation, and
  explicit errors for unsupported preservation combinations.
• Added a transform-to-JPEG benchmark profile covering quarter-turn and arbitrary rotation,
  crop-after-resize with a second resize, and combined flip/flop workflows, with isolated timing,
  peak RSS, dimensions, pinned output-pixel validation, and an equivalent-workflow Jimp comparison.
• Added a cross-codec transform contract that runs the full ordered transform chain from every
  supported decoder through JPEG output, plus focused arbitrary-rotation coverage for grayscale,
  RGB, and RGBA sample formats.
• Added a contributor guide covering tests, benchmarks, modular feature design, documentation,
  changelog entries, AI-assisted changes, MIT licensing, and conflict-free pull requests to main.
• Added a repository release skill with explicit authority, validation, packaging, publishing,
  provenance, and post-release consistency gates.
• Added a private vulnerability disclosure policy that defines the dependency-free runtime boundary
  while keeping CI, build, credential, and published-artifact supply-chain risks in scope.
• Added weekly and change-triggered CodeQL analysis for JavaScript and TypeScript using the extended
  security query suite.
• Added runtime validation for custom ImageSource reads, normalizing short, oversized, detached,
  and rejected reads into contextual ImageError results before codec parsing.
• Added a release-gated deterministic corruption campaign with hundreds of bit flips across a
  committed benchmark seed for every codec, saved raw-exception reproducers, and a checked-in
  regression corpus that automatically turns promoted crashes into permanent tests.

Changed

• Kept arbitrary-angle rotation memory bounded with a temporary 32x32 tile spool and small
  destination blocks, without per-pixel promises or temporary sample arrays.

PureJsImage v0.5.0

Added

• Added first-party ICO decoding with validated multi-image selection,
  zero-copy embedded PNG entries, common 1/4/8/16/24/32-bit DIB layouts,
  one-bit AND masks, partial alpha, and the Windows-compatible legacy
  all-zero-alpha fallback.
• Added a committed, checksum-pinned ICO corpus plus correctness-gated isolated
  benchmarks for metadata, embedded PNG and DIB decoding, masks, alpha, resize,
  PNG output, and JPEG output.
• Expanded first-party TIFF decoding with BigTIFF 64-bit IFDs and offsets,
  padded tiled layouts, 16-bit integer channels, CMYK and subsampled YCbCr
  conversion, CCITT Modified Huffman and Group 3 fax, and old/new JPEG-in-TIFF.