Why it matters
Without AI-guided colour matching, the risk of metamerism, irreversible colour changes, or incongruent fills increases, potentially delaying loans, provoking insurance disputes, and eroding client trust.
From design to safety
Design
AI informs treatment proposals
AI analyzes baseline colour data, original pigment references, ageing indicators, varnish layers, and expected glaze interactions to propose reversible pigments, compatible fills, and explicit colour targets with tolerances. It creates a treatment plan scaffold and a formal PACR-aligned proposal file for review.
- Spectrophotometer
- Digital Color Reference Library
- Retouching Records Software
Build
Materials and equipment alignment
Calibration routines align colour data feeds with reversible pigments, gap-fill media, and matching textiles, ensuring data-driven choices meet PACR ethics and conservation standards while documenting installation steps.
- Spectrophotometer
- Reversible Pigments Kit
- Gap-Fill Resin
Operation
Colour matching workflow in studio
AI calculates target colour, renders multiple inpainted swatches, compares to reference under D65 and UV, and documents rationale with on-screen validation prompts for conservator sign-off, including log of glaze depth changes.
- Spectrophotometer
- Gap-Fill Resin
- Retouching Records Software
Maintenance
Quality control and calibration
Regular grey-scale checks, reference panel scans, and spectral drift analysis validate colour consistency, triggering recalibration of sensors and updates to colour libraries and documentation templates, plus archival versioning.
- Digital Color Reference Library
- Retouching Records Software
- Portable Microscopy Station
Safety
Safety and ethics checks
AI flags irreversible choices, enforces reversibility constraints, and requires signed approval by the conservator before on-frame application and any on-site intervention, with COSHH-compliant PPE records.
- Retouching Records Software
- Portable Microscopy Station
- Digital Color Reference Library
Part by part
- 1
Spectrophotometer
- Watch for
- Metameric shifts under different gallery and daylight lighting conditions.
- AI action
- Reads spectral data and updates target colour space for match decisions.
- 2
Digital Color Reference Library
- Watch for
- Drift in reference swatches or inconsistent colour naming across sessions.
- AI action
- Normalizes readings, flags outliers, and aligns libraries with treatment goals.
- 3
Reversible Pigments Kit
- Watch for
- Degradation or incompatibility with original binding media.
- AI action
- Suggests compatible, reversible pigment choices and layer sequences.
- 4
Gap-Fill Resin
- Watch for
- Texture and translucency mismatch with surrounding paint.
- AI action
- Proposes fill opacity, texture parameters, and drying times to match adjacent areas.
- 5
Retouching Records Software
- Watch for
- Missing treatment rationale or missing colour target citations.
- AI action
- Generates standardised treatment report entries and linkage to condition records.
- 6
Portable Microscopy Station
- Watch for
- Inadequate surface cleanliness affecting colour readings.
- AI action
- Captures high-resolution images for documentation and cross-checks with colour data.
Worked example
Catching a metameric mismatch before a charter
Illustrative scenario with sample data.
A mid-19th century portrait shows subtle colour drift in skin tones under gallery lights; AI assists selecting a reversible colour system to correct without altering glaze depth.
Signal in
spectralData: {refPaint: [320-700nm], measuredPaint: [322-690nm]}, referenceLibrary: {pigments: ['lead white','azurite'], binder: 'oil'}, targetColor: '[CIE Lab: L=58 a=6 b=8]','gapFill: [opacity=0.2, texture=soft','lighting: 'D65'], ageingIndicators: ['oxidation','fatty acid migration'], conservationContext: 'loose varnish, overpaint present','treatmentConstraints: 'reversibility'
System out
inpaintPlan: {pigment: 'reversible aluminosilicate', fill: 'crackle-free', opacity: 0.22, margin: 0.5mm, notes: 'match to surrounding glaze'}, report: {treatmentProposal: true, neededConservationNotes: 'reversibility confirmed', approvals: 'conservator sign-off' }
Human sign-off: Head of Conservation reviews the AI-generated strategy and signs the final treatment proposal before execution.
How to roll it out
Phase 1
Weeks 1 to 4
Baseline colour data collection and library setup; establish reading routines and initial pigment matches.
Phase 2
Weeks 5 to 8
Pilot inpainting tests on mockups; calibrate AI colour decisions with conservator validation.
Phase 3
Weeks 9 to 12
Full integration into treatment workflow; documentation templates and archival records activated.
Where people stay in charge
- Decisions on irreversible changes must be made only by a qualified conservator with sign-off.
- All colour matching outputs must be reviewed against the documented treatment proposal by the head of conservation before execution.
- AI outputs cannot override documented professional ethics; human oversight required for all metamerism-related choices and provenance considerations.
Questions owners and crews ask
What data does the AI use for colour matching in painting restoration?
The AI uses spectral data, reference colour libraries, original pigment references, binder interactions, ageing indicators, varnish layers, and lighting conditions to propose reversible matches and documented decisions.
How are reversibility and ethics handled in AI-assisted inpainting?
The system flags irreversible options, enforces reversibility constraints, and requires a conservator’s approval, ensuring choices comply with minimal intervention and reversibility principles.
What records are created for AI-guided colour matching?
Treatment proposals, colour targets, pigment selections, fill parameters, inpaint plans, and a detailed retouching record are stored for condition reports and archives.
How is safety ensured during on-frame application?
Conservators validate AI guidance, employ PPE, adhere to COSHH controls, and ensure sign-off requirements are met before any on-frame action.