JLISwift v0.5.0

What's Changed

✨ Features

• lossless): restart intervals — decode, opt-in encode, parallel segments (P4b) (93aa5b0)
• bench): identity-hash gate + lossless clinical-path profiler (P0) (8f38fb4)

🐛 Bug Fixes

• build): explicit Foundation import for NSLock on watchOS (9f1c6eb)
• adversarial-review findings — heap over-read, segment validation, identity gaps (89999a4)

⚡ Performance

• encode): parallel lossless entropy emit via unstuffed-chunk stitch (P4a) (b31a23b)
• progressive) + chore: AC-pass buffer steal, remove dead Metal pipeline (P4c) (5260667)
• encode+decode): smaller hot-path items (P3c) (dab906d)
• decode): fused dequant+IDCT, parallel Step-7 reconstruction (P3b) (bbdb554)
• color): planar stride-1 vDSP conversion + vImage (de)interleave (P3a) (2b9c02c)
• dicom): cached W/L renderer, single-copy I/O, concurrent frame decode (P2b-P2d) (00577eb)
• encode): lossless residuals memoized once, swept in parallel (P1d) (e508e11)
• markers): readEntropyData as scan-then-slice (P2a) (b53b689)
• entropy): 64-bit BitReader with SWAR bulk refill (P1b) (16444bc)
• entropy): CLZ category + fused code+magnitude emit (P1a) (5e979d2)

🔧 Other Changes

• chore(release): bump version to 0.5.0 (ca4cc53)
• refactor(decode): lossless scan via raw-pointer static workers (P1c) (77f622f)

Installation

Add JLISwift to your Package.swift:

dependencies: [
    .package(url: "https://github.com/Raster-Lab/JLISwift.git", from: "0.5.0")
]

JLISwift v0.4.0

What's Changed

✨ Features

• decode): float32 output (raw sample values) for grayscale + XYB (WS-M4) (492ba26)
• dicom): multi-frame read/write + Basic Offset Table (WS-M2) (08e15a0)
• bench): --signed-stats medical evidence census (WS-M1) (0d1cc58)
• dicom): DICOM writer + JPEG-encapsulation join (eefbc54)
• safety): encode-side medical provenance + signed-pixel handling (9218a16)

🐛 Bug Fixes

• dicom): YBR_FULL/YBR_FULL_422 → RGB on render (WS-M2) (ed4479d)
• dicom): reader correctness + safety hardening (db3bf46)
• decoder): fail-safe guards for decompression bomb + lossless shift (196a852)

📚 Documentation

• roadmap): WS-M4 outcome — float32 built, perceptual items deferred/skipped (8e80679)
• regulatory): WS-M3 regulatory scaffolding drafts (7d0a681)
• roadmap): mark WS-M2 (real-clinical-data codec gaps) done (564fdd3)
• roadmap): mark WS-M1 (verification evidence pack) done (8dcc23c)
• roadmap): 0.4.0 plan — medical foundations (5605d0e)
• medical-grade assessment + remediation log (e42b888)

🔧 Other Changes

• chore(release): bump version to 0.4.0 (f50ae22)
• ci: medical conformance-evidence gate (WS-M1) (4d2e516)
• test(corpus): frozen content-addressed medical regression corpus (WS-M1) (061b6a5)
• test(evidence): medical conformance-evidence suite (WS-M1) (8111ecb)
• test(dicom): malformed-input fuzz harness for the DICOM reader (WS-M1) (8f690e6)
• test(dct): ISO/IEC 10918-2 Annex A IDCT accuracy conformance (4523864)

Installation

Add JLISwift to your Package.swift:

dependencies: [
    .package(url: "https://github.com/Raster-Lab/JLISwift.git", from: "0.4.0")
]

JLISwift v0.3.0

What's Changed

✨ Features

• lab): interactive DICOM window/level controls (a60bf19)
• quant): luma-derived field for chroma in jpegli AQ path (WS-A) (ab29815)
• bench,jlilab): batch metrics + DMG packaging scaffolding (WS-B) (8c5e806)
• quant): jpegli zero-bias adaptive-quant path, opt-in (WS-A part 2) (1038c6f)
• quant): jpegli adaptive-quant field computation (WS-A, 0.3.0) (bfb18e9)
• jlilab): add app icon (0bf483e)

📚 Documentation

• roadmap): re-scope 0.3.0 to the distributable signed app (df547f8)
• roadmap): WS-A adaptive-quant calibration finding (6bb6089)
• roadmap): 0.3.0 plan (jpegli adaptive-quant field + JLILab distribution) (4cae304)

🔧 Other Changes

• chore(release): bump version to 0.3.0 (8e8b3bf)

Installation

Add JLISwift to your Package.swift:

dependencies: [
    .package(url: "https://github.com/Raster-Lab/JLISwift.git", from: "0.3.0")
]

JLISwift v0.2.0

What's Changed

✨ Features

• encode)!: default to jpegli perceptual quant tables (WS4) (fd8cfe3)
• jlilab): WS1 rate–distortion curve view (2057b62)
• jlilab): WS1 honest measurement — SSIMULACRA2, spawn-correction, Debug banner (1b939e1)
• jlilab): cross-codec comparison + interop validation (cd44027)
• app): JLILab — DICOM/image round-trip quality lab (macOS) (0232151)

🐛 Bug Fixes

• jlilab): correct CT rendering (Modality LUT) and SwiftUI image display (abc040f)

⚡ Performance

• decode): parallelize chroma upsample across cores (WS3) (bead44a)
• decode): kill coefficient-write COW + scratch zero-fill (WS3) (dc620a0)
• decode): optimize chroma upsampling (WS3) (189be23)
• encode): skip zero-fill of fully-overwritten quantize buffers (WS2) (7d82109)
• color): cut color-conversion allocations + copy-on-write (WS2) (bb1aa01)

📚 Documentation

• roadmap): 0.2.0 optimization plan; mark 0.1.0 released (6699f96)

🔧 Other Changes

• test: assert version shape, not a literal (unbreak version bumps) (6916bbc)
• chore: bump version to 0.2.0 (d3f5ece)

Installation

Add JLISwift to your Package.swift:

dependencies: [
    .package(url: "https://github.com/Raster-Lab/JLISwift.git", from: "0.2.0")
]

JLISwift v0.1.0

What's Changed

✨ Features

• encode): accept float32 input (normalized [0,1] → 8-bit) (fa20201)
• encode): spatial adaptive quantization field (opt-in) (2e376d7)
• codec): XYB color encode + decode (experimental, opt-in) (90033db)
• color): XYB ICC profile generator (libjxl port), validated via CoreGraphics (bf3a681)
• color): correct XYB transform (faithful libjxl opsin port) (3a7fbf4)
• encode): restart markers in progressive scans (897ac18)
• codec): ICC profile + Exif metadata (embed on encode, extract on decode) (e5a5a58)
• encode): jpegli perceptual quantization tables (opt-in) (face4a4)
• codec): 1/2 & 1/4 scaled decode via exact low-frequency box-average (74f511d)
• codec): near-lossless SOF3 via point transform (encode + decode) (d804c16)

🐛 Bug Fixes

• encode): validate restartInterval against the 16-bit DRI field (22da0ac)

⚡ Performance

• decode): table-driven 8-bit-lookahead Huffman decode (338e1f0)
• xyb): enable per-channel trellis (RDO) for XYB encode (65b4d4b)
• encode): parallelize optimized-Huffman AC-frequency counting (011d195)
• encode): parallelize trellis quantization across cores (2495213)

📚 Documentation

• readme): bring up to date with the full feature set + honest caveats (6b2d888)
• accurate library overview + fix stale config docs/defaults (a2583c3)
• roadmap): XYB foundations done & validated (transform + ICC); wiring remains (6bb002e)
• roadmap): mark milestone complete — all large/marginal items done (e25e20c)

🔧 Other Changes

• test(fuzz): cover the new metadata + scaled-decode paths against malformed input (3229c1d)

Installation

Add JLISwift to your Package.swift:

dependencies: [
    .package(url: "https://github.com/Raster-Lab/JLISwift.git", from: "0.1.0")
]