Three formats have a serious claim to hold a scan for decades. The published guidance prefers two of them and tolerates the third, for reasons worth understanding before the first file is saved.
For a scan that is meant to outlast the scanner, the published guidance points to TIFF first, JPEG 2000 second and DNG as a tolerated third. The Federal Agencies Digital Guidelines Initiative names TIFF and JPEG 2000 as master formats for film. The Library of Congress prefers both and accepts DNG. The UK National Archives accepts lossless JPEG 2000 or TIFF. Nobody recommends JPEG for masters.
The reasons behind that ranking are more useful than the ranking. A preservation format has to do three things: hold every pixel value exactly, remain readable by software that has not been written yet, and carry enough description to identify itself. The three candidates meet those tests in different ways.
What makes a file format suitable for preservation?
The Library of Congress assesses formats against sustainability factors such as disclosure (is the specification public?), adoption (how widely is it used?) and freedom from patent constraints. Its Recommended Formats Statement for still images then adds technical preferences for the files themselves. As viewed on 17 November 2026, those were: the highest available resolution, not rescaled or interpolated; the highest available bit depth, 16 bits per channel if possible; an embedded colour profile; no compression, with lossless compression acceptable; and no layers.
That list is a compact specification for any master file, whichever container holds it.
Why is TIFF still the default?
TIFF is old, simple and everywhere. The Library of Congress format description dates the final TIFF 6.0 specification to 3 June 1992. It was developed by Aldus and Microsoft, and the specification now belongs to Adobe. It is fully documented, and the Library of Congress notes that particular subtypes of TIFF are widely used as master formats for scanned images and are supported by a great many applications.
For an archive, the practical merits are these.
- Pixels are stored plainly. An uncompressed TIFF is close to a raw grid of numbers with a header. If part of a file is damaged, the rest is often recoverable.
- 16-bit colour and embedded ICC profiles are standard.
- Embedded metadata is well supported. IPTC and XMP fields travel inside the file, and can be inspected with the EXIF and metadata viewer.
- Lossless compression is optional. The National Archives' guidance recommends ZIP compression and accepts LZW.
TIFF has two weaknesses. The files are large: a 6x7 centimetre transparency at 4,000 ppi and 16 bits is roughly 583 MB uncompressed, a figure you can reproduce for any format in the scan resolution calculator. And standard TIFF uses 32-bit offsets, which caps a file at 4 GB. Single-frame scans of roll film sit well below that, but stitched captures of large sheet film can approach it. BigTIFF, a variant with 64-bit offsets, removes the limit; if a project needs it, the brief should say so, because not all software reads it.
What does JPEG 2000 offer, and what is the catch?
JPEG 2000 is an international standard, ISO/IEC 15444, whose Part 1 defines the core coding system. It compresses with a wavelet transform and can work in two modes: lossless, where the decoded image is identical to the source, and lossy. It also allows a lower-resolution version of an image to be decoded from the same file without reading all of it, which suits very large images delivered over a network.
Its standing in guidance is strong. FADGI lists it beside TIFF for film masters. The Library of Congress lists JPEG 2000 (.jp2) among its preferred formats for photographs. The National Archives names JPEG 2000 with lossless compression first in its accepted formats.
The catch is the ecosystem. Fewer everyday applications open JPEG 2000 than open TIFF; it is telling that The National Archives' guidance points readers to a specific free viewer for the files. The mathematics is more complex, so a damaged file is harder to salvage. On patents, the Library of Congress records that the JPEG committee aimed for implementation free of royalties and licence fees and that several patent holders waived fees, while quoting a guide's warning that this is not a formal guarantee. And the extension gives no indication of whether lossless or lossy mode was used, so the setting has to be specified and verified.
For an institution with engineering support and many terabytes to store, lossless JPEG 2000 is a reasonable master format. For a small team that needs every tool to open every file, TIFF is the safer choice.
Where does DNG fit?
DNG, Digital Negative, is Adobe's openly published format for camera raw data. Adobe describes it as a publicly available archival format for raw files and as an extension of TIFF 6.0, with a patent licence offered to implementers without royalties. On Adobe's page the current specification was version 1.7.1.0, dated September 2023, when checked on 17 November 2026.
DNG matters to film archives because of camera scanning. A camera pointed at a negative produces raw sensor data, and that data has not yet been turned into a finished RGB image. Keeping it as DNG instead of a manufacturer's proprietary raw format replaces an undocumented container with a documented one.
But raw data is not a picture. It has to be rendered, and different software, or the same software in a later version, will render it differently. A raw capture of a colour negative is still a negative with its orange mask; the positive exists only after conversion, a step the negative to positive converter demonstrates on a small scale. That dependence on rendering is the practical difference from the other two formats. The Library of Congress places DNG in its acceptable list for photographs, not its preferred one, and FADGI's film tables name only TIFF and JPEG 2000.
A common arrangement for camera-based projects is therefore to keep both: the DNG as the capture record, and a rendered 16-bit TIFF as the preservation master that people will actually use.
How do the three formats compare?
| TIFF | JPEG 2000 (lossless) | DNG | |
|---|---|---|---|
| Holds | Rendered image | Rendered image | Raw sensor data, usually with a preview |
| Published by | Adobe (originally Aldus) | ISO/IEC 15444 | Adobe |
| FADGI film master format | Yes | Yes | Not listed |
| Library of Congress, photographs | Preferred | Preferred | Acceptable |
| UK National Archives guidance | Accepted (baseline TIFF 6.0) | Accepted (lossless) | Not mentioned |
| File size | Largest | Smaller | Smaller than a rendered TIFF |
| Software support | Very wide | Narrower | Wide in photographic software |
| Main caution | 4 GB cap without BigTIFF | Verify lossless mode; fewer tools | Needs rendering; keep a TIFF too |
What about JPEG, PNG and everything else?
JPEG is for access. It is lossy and 8-bit, and it is what picture desks, websites and email expect. National Lottery Heritage Fund guidance draws the line clearly: TIFF for archival masters, JPEG for access copies. Make JPEGs from the master, never the other way round; the TIFF to JPG converter does this for 8-bit and 16-bit scans in the browser, with the limitation that embedded metadata and colour profiles are not carried into the output.
PNG is lossless and appears on the Library of Congress preferred list, but it is rarely used for photographic masters. Layered editing files and proprietary raw files are working formats, not preservation ones.
Three further settings matter as much as the container.
- Bit depth and colour profile. Sixteen bits per channel with an embedded profile, as covered in our guide to scanning medium-format transparencies.
- Fixity. Record a SHA-256 checksum for each master when it is created. The image hash and fingerprint tool generates one for a single file; at volume, the same value should be written to the catalogue automatically.
- Storage arithmetic. Format choice multiplies through every copy. The digitisation storage calculator totals masters and derivatives with a 3-2-1 backup multiplier.
Then write the decision down, in one paragraph, in the project specification. A consistent, documented choice of any of these formats will serve an archive better than a superior format applied unevenly.
The decision ahead for the Robert Harper archive
The Robert Harper Photograph Library holds 52,684 analogue photographs by its inventory, alongside 42,803 digital photographs, and the analogue material largely awaits scanning. Its master format has not been fixed, and it will be settled with whoever carries out the capture. Imaging vendors and digital preservation specialists with a view on that choice are invited to say so through the digitisation partners page.
Frequently asked questions
What is the best file format for archiving scanned photographs?
TIFF is the most widely accepted. FADGI lists TIFF and JPEG 2000 as master formats for film, the Library of Congress prefers both for photographs, and the UK National Archives accepts either. TIFF's advantages are simplicity and near-universal software support. JPEG 2000 with lossless compression saves storage at the cost of narrower tool support.
Is DNG a good archival format?
It is an acceptable one, with a specific use. DNG is Adobe's openly documented raw format, an extension of TIFF 6.0. The Library of Congress lists it as acceptable, not preferred, for photographs. For camera-based film capture it preserves the sensor data, but it needs rendering software to become a viewable picture, so most archives also keep a rendered TIFF.
Is lossless JPEG 2000 really lossless?
Yes, when the lossless mode is used. JPEG 2000 can compress either reversibly, so that the decoded pixels are identical to the original, or lossily. The file extension does not tell you which was chosen, so a specification must say lossless explicitly and the vendor's settings should be checked on sample files.
Should I compress TIFF files?
Lossless compression does not alter pixel values. The UK National Archives recommends ZIP compression for TIFF and accepts LZW. The Library of Congress states a preference for uncompressed files and treats lossless compression as acceptable. Uncompressed TIFF is the simplest to recover if a file is damaged; compressed TIFF saves storage. Either is defensible if applied consistently.
Can JPEG be used as a preservation format?
Not for scan masters. JPEG discards data each time it is saved and is limited to 8 bits per channel. It is the right choice for access copies, previews and web delivery. National Lottery Heritage Fund guidance puts it simply: uncompressed TIFF for archival masters, JPEG for compressed access copies.
Sources
- FADGI: Technical Guidelines for Digitizing Cultural Heritage Materials, third edition (May 2023)
- Library of Congress: Recommended Formats Statement, still image works
- Library of Congress format description: TIFF, Revision 6.0
- Library of Congress format description: JPEG 2000 Part 1, Core Coding System
- Library of Congress format description: Adobe Digital Negative (DNG), Version 1.6
- Adobe: Digital Negative (DNG)
- The National Archives: Digitisation (guidance on digitising records)
- National Lottery Heritage Fund: Doing digitisation on a budget
Published by Archive.London. Corrections are welcome under our editorial policy.

