Cultural heritage carries the memory of human civilization, but time, environment, and human factors are gradually causing these memories to fade. Restoration and conservation are no longer just the work of hands and brushes — they are becoming a precision data engineering discipline.
Among all categories of cultural heritage, manuscripts and rare books impose the most stringent requirements on digitization technology due to their fragility and irreplaceability. A yellowed page may carry thoughts from a thousand years ago; a brittle scroll may crumble the moment it is opened. How to completely transform every ink stroke, every texture, every layer of hidden information into digital form without touching or damaging the original — this is the ultimate challenge of manuscript digitization, and the core application scenario of industrial imaging technology.
This article takes manuscript and rare book digitization as its main thread, outlining the technical logic, industry practices, and selection framework for reference by practitioners and project planners in related fields.

The demand for manuscript digitization is growing at an astonishing rate. The global rare book digitization market is expected to reach approximately $15 billion by 2025. In China, with the deepening of the national cultural digitization strategy, the market size related to rare book digitization protection exceeded 5 billion yuan in 2023 and is expected to reach over 12 billion yuan by 2026. By January 2025, a total of 158,000 rare book digital resources had been released online nationwide.
But this expansion also brings new challenges: how to maintain consistent image quality across massive scanning operations? How to avoid damaging fragile originals while maintaining high-speed capture? How to make digital surrogates not just "visible," but "clear" and "deep"? Consumer-grade equipment cannot answer these questions.
Digitizing a precious rare book is far more than "taking a photo." According to the "Technical Specifications for Rare Book Digitization," manuscript scanning must meet the following requirements: single-page image pixels no less than 2400×2400, grayscale levels no less than 256, and color reproduction error controlled within ΔE<3. During image acquisition, environmental parameters such as illumination intensity (500-1000 lux) and color temperature (5000K±200K) must be strictly controlled.
More importantly, the materials of rare books vary widely — from hemp paper of the Tang and Song dynasties to bamboo paper of the Ming and Qing, from parchment to palm-leaf manuscripts, from ink writings to cinnabar annotations — each medium imposes different requirements on imaging. This requires imaging systems to have adjustable spectral response, precise color management, and flexible lens configuration capabilities.

Unlike traditional area scan cameras that capture a complete image in one shot, line scan cameras use one-dimensional linear array sensors to scan the target line by line, then stitch thousands of lines of image data into a complete two-dimensional image.
Time Delay Integration (TDI) is a key technology in line scan cameras. It uses multiple rows of sensors to expose the same target multiple times and accumulate signals, significantly improving sensitivity and signal-to-noise ratio. The higher the TDI stage (e.g., 64-stage, 128-stage, 256-stage), the better the imaging quality under low-light conditions — which is particularly important for scenarios where illumination needs to be reduced to protect fragile manuscripts.
A vast amount of information in manuscripts is invisible to the naked eye — text covered by stains, faded annotations, erased or overwritten records, superimposed ink from different periods. Multispectral imaging technology captures images across different wavelength bands, revealing the information layers of a document one by one.
Different materials have different absorption, reflection, and fluorescence characteristics for different wavelengths of light. By capturing images in ultraviolet (UV, 200-400nm), visible (VIS, 400-700nm), near-infrared (NIR, 700-1100nm), and short-wave infrared (SWIR, 900-1700nm) bands respectively, the information layers of a document can be revealed layer by layer:
In the "digital twin" scenario for painting restoration, high-resolution area scan cameras play an important role. Conservators need to pre-simulate restoration plans on a multi-gigapixel digital copy, which requires the imaging system to have extremely high spatial resolution and color reproduction accuracy.

This technology has been widely adopted in restoration projects at top museums and auction houses worldwide, becoming one of the standard tools for painting conservation.
Manuscript digitization is evolving from simple "flat reproduction" to multi-dimensional information acquisition systems:
The British Library is currently conserving and digitising nearly 800 Chinese Lotus Sutra manuscripts, making them freely available on the International Dunhuang Project (IDP) website. The manuscripts were discovered in a hidden cave in Dunhuang in 1900.
The Vatican Apostolic Library has planned to digitize all the manuscripts in its collections. The project uses V-shaped glass lecterns suitable for scanning books or documents that cannot be fully opened 180 degrees, perfectly aligning with the non-contact working method of line scan cameras. The project places special emphasis on protecting fragile materials and is regarded as one of the global benchmarks for manuscript digitization.
Since the early 1990s, the Dunhuang Academy has launched the "Digital Dunhuang" project. By the end of 2023, the team had completed digital acquisition of murals from 295 caves and digitized scanning of 50,000 historical archive negatives. In 2025, the National Library of France donated its entire digital copies of Dunhuang manuscripts to Peking University — marking a new phase in international sharing of digitized manuscript resources.
The digitization of the Wensuge Siku Quanshu is one of the largest-scale manuscript digitization projects in China. The first phase was completed in June 2022, comprising digitized acquisition of 6,141 cases, 3,474 titles, 36,315 volumes, and over 2.44 million pages — all scanned at original size with high-definition imaging.
The resolution requirements for manuscript digitization are far higher than for ordinary document scanning. According to industry standards, optical resolution should not be lower than 600 dpi. For A2 format (420mm×594mm), 600 dpi corresponds to approximately 9,900×14,000 pixels — about 140 megapixels — far exceeding the single-shot capability of any consumer camera.
Therefore, manuscript digitization typically adopts two approaches:
The color information of manuscripts — the degree of paper yellowing, the density variations of ink, the vividness of cinnabar annotations — is itself important scholarly information. According to specifications, color reproduction error should be controlled within ΔE<3.
The primary principle of manuscript digitization is "do no harm to the original." This includes:
Large-scale manuscript digitization projects often involve millions of pages of scanning work. Taking the Wensuge Siku Quanshu as an example, the first phase alone completed acquisition of over 2.44 million pages. At this scale, scanning speed directly determines project timelines and costs.
Line scan cameras can achieve efficient scanning while maintaining 600 dpi resolution — for example, an 8K line scan camera can complete several A2-format pages per second. But speed improvement must not come at the cost of image quality — this is the fundamental difference between industrial-grade and consumer-grade equipment.
A complete international standards system has been established in the field of manuscript digitization:
| Standard | Issuing Body / Region | Core Requirements |
|---|---|---|
| FADGI | U.S. Federal Agencies | Technical benchmarks for color accuracy, resolution, and file formats |
| Metamorfoze | Netherlands | ≥400 dpi with color management |
| ISO 19264-1 | International Organization for Standardization | Digital imaging quality metrics |
| GB/T 37069-2018 | China | Resolution, grayscale levels, and color reproduction for manuscript scanning |
| Dublin Core | International | Metadata description standard |
In practice, quality control runs through the entire workflow:
The following is a reference for some industrial camera product types available on the market related to manuscript digitization scenarios:
| Product Series | Model | Key Features | Recommended Application |
|---|---|---|---|
| 10GigE Line Scan | MV-XGL83M | 8K · 105kHz line rate · 2Line TDI · 10GigE | Primary choice for large-scale manuscript digitization |
| MV-XGL162M | 16K · 72kHz line rate · 256-stage TDI · 10GigE | Ultra-fine detail capture | |
| CXP High-Speed Line Scan | MV-PXL90M | 9K · 529kHz line rate · 256+32-stage TDI · CXP-12 | Ultra-high-speed large-scale digitization |
| MV-PXL164BM | 16K · 56kHz line rate · 2TDI mode · CXP-12 | Large-format high-precision scanning | |
| CXP High-Resolution Area Scan | MV-PX2500GC/M | 25MP · global shutter · 150fps · CXP-12 | Painting digital twin detail acquisition |
| MV-PX41000C/M | 410MP · 24K · TEC+liquid cooling · CXP-12 | Large-format painting ultra-high-precision imaging | |
| UV/SWIR Multispectral | MV-PXL90U | 9K · 250-380nm UV-enhanced · CXP-12 | Fluorescence detection, restoration trace ID |
| MV-GEC500I | 5MP · 0.4~1.7μm broad spectrum · USB3.0 | Ink composition analysis, pigment ID | |
| Thermal Camera | MV-GF480 | 0.48MP · 800×600 · 8-14μm · ±1℃ accuracy | Storage environment monitoring, preventive conservation |
| Smart Camera | MV-ITA134GC-T | X86 architecture · built-in AI inference · 5MP | On-site AI-assisted authentication |
A: Line scan is suitable for large-scale, standardized scanning — continuous operation, no stitching seams, and the only viable solution for scrolls and large-format materials. Area scan is more flexible, suitable for manuscripts with special bindings that cannot be laid flat. The two can also be used in combination — line scan for large-scale text scanning, and area scan for covers, title pages, special inserts, and other parts that require flexible composition.
A: For most projects, yes — this is the minimum requirement of international standards such as FADGI and Metamorfoze. For paper fiber analysis, ink penetration research, or AI restoration of severely damaged manuscripts, 800-1200 dpi is recommended. The trade-off is: every doubling of resolution increases image data volume by four times, raising requirements for storage, transmission, and processing capabilities.
A: Not for every project. Strongly recommended for: documents with stains covering text, signs of erasure or alteration, ink from different periods superimposed, and materials requiring pigment differentiation. If manuscripts are in good condition and only readable copies are needed, visible light imaging suffices. Multispectral acquisition can be conducted as a supplement to key documents after basic digitization, but completing it in one go costs far less than rework.
A: LED cold light sources are the top choice — low heat, long lifespan, stable spectrum. Recommended color temperature: 5000K±200K (D50 standard). Illumination: 500-1000 lux, below 500 lux for fragile materials. Color Rendering Index (CRI) should be above 90, preferably ≥95. Also avoid glare and reflection — glossy paper surfaces require polarizers or diffused light sources.
A: GigE (1 Gbps): 100m cable, low cost, suitable for conventional projects. 10GigE (10 Gbps): 10× bandwidth of GigE, suitable for 8K-16K high-speed scanning — the most balanced choice for most manuscript projects. CoaXPress (up to 12.5 Gbps): ultra-high speed, extremely low latency, suitable for 16K+ or top-tier projects with extreme speed requirements.
A: Beyond the equipment purchase, a significant portion of the budget should be reserved for: ① dedicated line scan lenses (ordinary lenses cannot be used); ② high-quality lighting systems; ③ mass data storage and high-speed transmission solutions; ④ development and configuration of scanning software and color management systems; ⑤ operator training and regular maintenance/calibration. It is recommended to include these in the overall budget planning at the project initiation stage.
Manuscript digitization is becoming a comprehensive discipline integrating precision optics, multispectral analysis, artificial intelligence, and knowledge engineering. The starting point of all this is the industrial camera running steadily on the scanning table — faithfully recording the spectral information of every ink stroke in 0s and 1s, resisting the erosion of time through line-by-line scanning.
For a library or cultural heritage institution, a good digitization imaging system is not simply equipment procurement — it is a bridge: connecting fragile originals with eternal digital surrogates, connecting the wisdom of the past with the readers of the future. Millennium-old texts may fade, but their digitized versions will not.
This article aims to provide an objective and practical visual technology reference for digitization and restoration projects in the cultural heritage field. For technical exchange, equipment testing, or solution evaluation on specific projects, please feel free to contact the MindVision team.
MindVision delivers industrial-grade machine vision cameras for quality inspection and factory automation — featuring global-shutter CMOS sensors, flexible interfaces (GigE Vision / USB3 Vision / CoaXPress-12), and the resolution and speed demanded by modern production lines. Share your part tolerance, line speed, and defect specifications — we'll help you select the right camera and avoid the accuracy pitfalls outlined in this guide.
Please Login