CCMX — Camera Colorimeter Correction Database / 相机色度计校正数据库

A community, open database of CCMX-style 3×3 correction matrices that turn a camera into a colorimeter for display measurement. Each correction maps a specific camera's raw RGB to accurate CIE XYZ for a specific display, keyed by (camera device × lens × display).

一个社区开源的 CCMX 风格 3×3 校正矩阵 数据库，用于把相机当作色度计来测量显示器。 每条校正把某相机的原始 RGB 校正为某显示器下准确的 CIE XYZ，按 (相机设备 × 镜头 × 显示器) 区分。

Tool-agnostic. Any application that uses a camera as a colorimeter can consume these corrections — the format below is self-contained. 与工具无关。任何"用相机当色度计"的应用都可使用这些校正——下述格式自包含。

📷 Scope: fixed-lens cameras only / 适用范围：仅固定镜头相机 A correction is bound to the camera's spectral response. This is stable only for fixed-lens cameras (e.g. phone/tablet cameras), where a per-(model × lens × display) correction is meaningful and shareable. Interchangeable-lens cameras (DSLR/mirrorless) are generally out of scope: each lens alters the transmission spectrum, so a single matrix is not portable across lens/body combinations. 校正与相机的光谱响应绑定，只有固定镜头相机（如手机/平板摄像头）才稳定，按 (机型 × 镜头 × 显示器) 的校正才有意义、可共享。可换镜头相机（单反/微单）一般不在 接收范围：每支镜头都会改变透过光谱，单一矩阵无法在不同镜头/机身组合间通用。

---

Why / 为什么

A camera's spectral response differs from the CIE standard observer and varies by model, and it reacts differently to each display technology (WOLED / QD-OLED / IPS …). A single generic matrix is not accurate enough, so corrections are calibrated per device and display against a trusted reference instrument (spectroradiometer / colorimeter).

相机的光谱响应不同于 CIE 标准观察者、且因机型而异，并对不同显示技术（WOLED / QD-OLED / IPS …） 表现不同。单一通用矩阵不够准，因此需对参考仪器（光谱仪 / 色度计）按设备与显示分别标定。

⚠️ A 3×3 matrix works for additive displays. For non-additive panels (WRGB OLED, QD-OLED) it is only an approximation and is flagged nonAdditiveWarning=true; those are better served by spectral (CCSS) corrections. 3×3 矩阵对加性显示有效；对非加性面板仅为近似，会标记 nonAdditiveWarning=true。

---

File naming / 文件命名

{deviceIdentifier}_{lensID}_{displayID}.json
e.g.  iPhone17,1_main_PA32UCG.json

• deviceIdentifier — camera device hardware id (e.g. Apple uname → iPhone17,1).
• lensID — main / ultrawide / telephoto / front.
• displayID — identifier of the calibrated display.

---

Format / 格式

Each file is a JSON bundle. The matrix maps camera linear RGB → absolute XYZ: XYZ ≈ M · max(cameraRGB − blackLevel, 0).

{
  "version": "1.0",
  "deviceIdentifier": "iPhone17,1",
  "deviceModel": "iPhone 16 Pro",
  "calibrationDate": "2026-06-18T10:00:00Z",
  "display": {
    "id": "PA32UCG",
    "name": "ASUS ProArt PA32UCG",
    "manufacturer": "ASUS",
    "displayType": "miniLED"
  },
  "matrices": {
    "main": {
      "row0": [484.37, 78.90, 135.76],
      "row1": [144.97, 335.40, -19.89],
      "row2": [5.59, -107.75, 896.88],
      "cameraIdentifier": "iPhone17,1_main",
      "displayID": "PA32UCG",
      "lensID": "main",
      "calibrationDate": "2026-06-18T10:00:00Z",
      "calibrationISO": 54.0,
      "calibrationExposure": 0.0126953125,
      "validation": { "deltaEMean": 0.67, "deltaEMax": 2.03, "patchCount": 24 },
      "blackLevel": { "offset": [2111.7, 2111.7, 2111.6], "enabled": true }
    }
  },
  "metadata": {
    "displayTechnology": "miniLED",
    "whitePointXY": [0.3127, 0.3290],
    "referenceInstrument": "i1Display Pro",
    "author": "your-handle",
    "license": "CC0",
    "nonAdditiveWarning": false
  }
}

Field notes / 字段说明

• matrices is keyed by lensID; each entry is one 3×3 (row0/row1/row2) plus optional blackLevel, validation (CIEDE2000) and capture settings.
• metadata is optional & backward-compatible (older files omit it). displayTechnology ∈ woled | qdOLED | oled | ipsWhiteLED | ipsWideGamut | va | miniLED | other.

---

Contribute / 如何贡献

1. Measure patches on the target display with a reference instrument and fit a 3×3 (least squares; validate with CIEDE2000). 用参考仪器逐块测量目标显示并拟合 3×3（最小二乘；CIEDE2000 验证）。
2. Save a JSON file named per the convention above and fill metadata (author, license — preferably CC0 —, referenceInstrument, displayTechnology). 按规范命名并填写 metadata。
3. Open a Pull Request. 提交 Pull Request。

Submissions are generally accepted only for fixed-lens cameras (e.g. phone cameras). Interchangeable-lens (DSLR/mirrorless) data is out of scope — see Scope above. 一般只接受固定镜头相机（如手机摄像头）的数据；可换镜头（单反/微单）不在范围内，见上文适用范围。

Some apps offer an in-app acquisition flow that exports a ready-to-submit JSON. 一些 app 提供应用内采集流程，可直接导出符合规范的 JSON。

---

Consume / 如何使用

Clone or pull this repo and read the JSON files; map the fields into your own pipeline. 克隆/拉取本仓库并读取 JSON，将字段映射进你自己的流程。

git clone <this-repo-url>
# corrections are flat *.json files at the repo root, named {device}_{lens}_{display}.json
# 校正为仓库根目录的扁平 *.json 文件，命名为 {设备}_{镜头}_{显示}.json

---

License / 许可

Unless a file's metadata.license states otherwise, corrections here are dedicated to the public domain under CC0 1.0. See LICENSE. 除非文件 metadata.license 另有声明，本仓库校正数据以 CC0 1.0 公共领域奉献，见 LICENSE。