Scratched and mould-damaged negatives UK: what a Coolscan 9000 ICE pass can still recover

Digital ICE on the Nikon Super Coolscan 9000 ED recovers the great majority of physical damage on a 35mm or 120 colour negative — emulsion-side scratches, surface dust, dried fingerprint smudges and clean-edged tears all disappear cleanly because the scanner's fourth infrared channel reads them as opaque and reconstructs the affected pixels from neighbouring data. On a typical scratched UK family C-41 negative (Kodak Gold, Fujicolor, Agfa Vista) the visible improvement is dramatic: a 1986 holiday roll that arrives at the lab with twenty visible white emulsion scratches comes back with none. But there are three film types where ICE quietly produces a worse result than ICE-off + a manual clean, and a fourth where it works but masks damage the customer needed to see. This article explains the mechanism, names the failure cases on real UK family-archive film stocks, and gives a 4-step triage you can run on a stack of negatives before booking a service.

By Maria C — Media Preservation and Heritage Specialist, EachMoment lab, Surrey. Twelve years working on UK family-archive negatives across the four Coolscan 9000 EDs the lab keeps in rotation. The opinions in this article come from operator log notes, not marketing copy.

What Digital ICE actually does (and why it is not magic)

Digital ICE — Image Correction and Enhancement — is a hardware-plus-software combination developed by the Kodak Austin Development Center in the late 1990s and licensed to scanner manufacturers including Nikon, Minolta, Canon and Epson. On the Coolscan 9000 ED it appears in the LaserSoft SilverFast or Nikon Scan menu as ICE, ICE Pro (four-channel) or ICE Pro Brilliance. The Pro tier is what the lab uses; the consumer "ICE" tier on cheaper scanners often refers to a software-only dust-removal step with no hardware infrared component.

The mechanism is genuinely clever and almost trivially simple. The scanner has the usual red, green, blue channels — and a fourth channel using infrared light at around 880 nanometres. At that wavelength the dye-cloud emulsion of a C-41 colour negative or E-6 colour slide is effectively transparent: the dyes were engineered to absorb visible light only. But anything physically sitting on or in the emulsion — dust, scratches, fingerprint oils, a hair, a strip of tape — blocks infrared just as it blocks visible light. The software diffs the IR channel against the RGB channels, identifies pixels where IR shows an obstruction but RGB shows image information, and infills those pixels from the surrounding image data.

This is the whole trick. It is not AI restoration. It is not generative. It is an infrared mask plus a content-aware fill, and the quality of the result depends entirely on whether the IR-transparent-emulsion assumption holds for your particular film.

When ICE works (and what the UK SERP misses about this)

The competing UK pages on this query — pixave.co.uk, filmscanuk.co.uk, negativefilmscanning.co.uk, takeiteasylab.com, mrscan.co.uk, revivestudios.co.uk — all mention "Digital ICE removes dust and scratches" or words to that effect as a single line of marketing copy on a pricing page. None explain when it fails. That is the gap this article fills.

On the films where ICE works it is genuinely close to magic. The infrared assumption holds cleanly on:

• Kodak Gold 100/200/400, Kodacolor, Kodak Royal Gold — the dominant UK consumer C-41 stock from roughly 1986 onward. Most family negative boxes from the 1990s are 80% Kodak Gold.
• Fujicolor Superia, Fujicolor 200, Fujicolor C200 — Fuji's UK consumer C-41 line. Slightly different dye-cloud chemistry but the IR assumption holds equally well.
• Agfacolor, Agfa Vista, Agfa Optima — German C-41 stocks present in UK family archives via Boots Photo and Truprint mail-order processing through the 1990s.
• Kodak Ektachrome (E-6), Fuji Provia (E-6), Fuji Velvia (E-6) — colour slide film. Behaves the same as C-41 under infrared because the emulsion is dye-cloud.
• Modern chromogenic black-and-white like Ilford XP2 Super and Kodak BW400CN — these are processed in C-41 and have no silver image, only dye clouds, so they pass the IR test cleanly.

If the box of negatives on your kitchen table is mostly orange-masked colour film with a "C-41" or "Process C41" stamp on the leader, ICE Pro on the Coolscan 9000 ED will solve almost every dust and scratch problem you have. The question only matters when the box has other things in it as well.

The three cases where ICE quietly makes the scan worse

These are not edge cases. Across UK family archives of negatives from the 1950s-1990s, roughly one box in three contains at least one of these — usually mixed in with the colour negatives so the operator has to triage by film type before scanning.

Case 1: traditional silver-based black-and-white negative film

This is the largest category of ICE failure in UK family archives. Pre-1990 black-and-white negatives — Ilford HP5, Ilford FP4, Ilford Pan F, Kodak Tri-X, Kodak Plus-X, Kodak Verichrome Pan, Agfa APX — develop a metallic silver image, not a dye-cloud image. The silver halide crystals are reduced to elemental silver during development, and elemental silver is opaque to infrared at the wavelengths Digital ICE uses.

The infrared channel cannot tell the difference between a silver grain (real image data) and a dust speck (a defect). Both block IR. ICE treats both as a defect, masks them, and "fills" them from surrounding pixels. The visible result on a Tri-X negative is a print that looks soft-but-clean instead of sharp-and-grainy: the characteristic Tri-X grain structure is gone, fine highlight detail (a white shirt collar, a printed road sign, hair texture in a portrait) collapses to mush, and the operator is left with a file that looks like a heavy-handed denoise pass. The damage is irreversible because the original grain information has been thrown away.

If a UK family-history client sends in a 1962 wedding album with Verichrome Pan rolls in it, ICE-on output looks superficially fine to a non-photographer but is unusable as an archival master. We always scan silver-image black-and-white twice: once with ICE off + multi-sample 16× as the master, once with ICE on as a "social media preview" if the client wants visible-defect-free reading copies.

Case 2: classic Kodachrome (K-14 process, 1936-2010)

Kodachrome is the famous edge case because it confuses people: it looks like colour slide film, but its emulsion chemistry is closer to silver-image black-and-white than to E-6 colour. The K-14 process forms dye images around retained silver, and a finished Kodachrome slide retains enough silver in its emulsion stack to block infrared. ICE on Kodachrome produces a softened, posterised mess. Nikon explicitly documented the incompatibility in the Coolscan 9000 ED manual and offered a separate "Kodachrome" mode that bypasses ICE and uses a different colour profile.

Kodachrome was sold in the UK from the 1930s through the early 2000s and remains the dominant slide stock in UK family archives covering 1955-1985 — particularly anything that came back from Switzerland or the United States in a yellow Kodak mailer. The Coolscan 9000 ED has the right software path for Kodachrome (Kodachrome-specific colour profile, ICE off, multi-sample 16×). A flatbed scanner without that path will scan the dyes correctly but cannot avoid the ICE-induced silver smear if the operator leaves ICE on.

Case 3: mould-damaged emulsion (any film stock)

The third case is not a film type but a damage type. UK family negatives stored in loft or garage spaces — temperature swings, summer damp, no climate control — frequently develop biological damage in the emulsion. Fungal hyphae from common household moulds (Aspergillus, Cladosporium, Penicillium are the usual suspects) feed on the gelatin binder, leaving star-shaped bloom patterns visible under raking light, often accompanied by a faint musty smell when the strip is held to the nose.

This is where the ICE failure mode is most dangerous because the output looks superficially acceptable. Mould damage is image loss — the emulsion under the bloom is partially eaten and the image information there is genuinely gone. ICE has nothing to interpolate from. Its content-aware fill smears whatever is around the damaged area across the bloom, producing a soft fuzzy patch that looks like a slight focus problem rather than what it really is: a hole in the historical record. A scan with ICE off shows the bloom honestly; the customer can then decide whether to commission a wet-clean (we use PEC-12 film cleaner with cotton swabs under microscope inspection) before re-scanning. ICE-on conceals the problem and the customer never knows.

A fourth case worth mentioning: high-density Dmax fail

There is a fourth, narrower failure case worth flagging. Very dense parts of a negative — under-exposed deep shadows on a 1970s available-light interior shot, or the maximum-density "Dmax" patches on a heavily-pushed Tri-X roll — can themselves block enough infrared that ICE mistakes them for a defect. The result is small irregular dark patches "corrected" out of existence in the deepest shadows. The Coolscan 9000 ED's Dmax of 4.8 is high enough that this only happens on genuinely extreme negatives, but it is the reason the SilverFast software ships with a separate iSRD ("Infrared Smart Removal of Defects") tier with adjustable sensitivity — at the lab we drop the iSRD sensitivity slider to 30/100 on dense negatives and rely on a manual cleanup pass for the rest.

Why the Coolscan 9000 ED specifically

The Coolscan 9000 ED is genuinely an unusual machine to still be using in 2026. Nikon discontinued it in 2009, never released a successor, and the active drivers are now community-maintained through the open-source VueScan and the commercial SilverFast 9. The reason UK and European labs still buy one on the secondhand market — usually £2,400 to £4,200 in working condition with original holders — is that the combination of features below has never been replicated in a single subsequent device.

Nikon Super Coolscan 9000 ED

UK lab reference — 35mm and 120/220 + ICE Pro

2003-2009 (discontinued; mature secondhand market)

• Hardware four-channel infrared — true Digital ICE Pro
• Optical: 4,000 DPI native — 3,900 measured (filmscanner.info)
• FH-869 medium-format holder; fluid-mount compatible
• Dmax 4.8 — recovers shadow detail flatbeds crush
• Multi-Sample scanning up to 16× for low-noise output
• Kodachrome-specific software profile (ICE off, K-14 colour)
• UK secondhand May 2026: £2,400-£4,200 with original holders

Plustek OpticFilm 120 Pro

Current production — software ICE only

2017-2026 (still manufactured)

• Software-only Digital ICE equivalent (no hardware IR channel)
• Optical: 5,300 DPI advertised — ~2,800 measured
• Accepts 35mm strips, mounted slides, 120 strips up to 6×12 cm
• Dmax 4.0 manufacturer spec — softens in deep shadows
• UK new May 2026: £1,899 RRP

Epson Perfection V850 Pro

Flatbed — strong on glass and 4×5", weak on roll-film

2014-present

• Has Digital ICE Professional via SilverFast Ai Studio
• Optical: 6,400 DPI advertised — 2,300 measured on 35mm
• Dmax 4.0 — IR channel less sensitive than Coolscan
• Genuinely useful for 4×5" sheets and glass plates
• UK new May 2026: £899 RRP

Reflecta ProScan 10T

Dedicated 35mm — Magic Touch dust removal

2019-2026

• Magic Touch (LaserSoft) — software dust/scratch correction
• Optical: 10,000 DPI advertised — ~3,000 measured
• 35mm only — no medium-format path
• UK new May 2026: £379 RRP

DSLR macro-rig (Nikon Z7 II + ES-2)

Backup chain when Coolscans are queued

2018-2026

• No infrared dust channel — ICE not available at all
• Effective 35mm resolution: ~4,200 DPI from 45.7 Mpx sensor
• Speed: ~8 seconds per frame — fastest path at high resolution
• Manual dust cleanup required in post for every frame

High-street flatbed under £200

What most UK competitor pricing pages actually use

Various — Canon CanoScan, Epson V370/V550

• No hardware IR — "ICE" advertised is software only
• Effective DPI on 35mm: ~1,200-1,800 measured
• Dmax 3.4 — shadows crushed in dense negatives
• UK new May 2026: £79-£199

When ICE helps vs harms by film stock

A 4-step triage to run on a box of negatives before booking a service

If you have a stack of family negatives in front of you and you are deciding whether to send them out, run these four checks before you contact any UK lab. They take about five minutes for an average shoebox and will tell you whether "yes please turn on Digital ICE" or "please discuss with me before scanning" is the right brief.

1. Sort by colour vs black-and-white. Hold strips up to a window. Colour negatives have a uniform orange mask across the entire film base; black-and-white negatives are a clear or pale grey base. Separate the two piles.
2. Check the leader and edge text. The edge of every roll should be printed with the film stock name. KODAK GOLD 200, FUJICOLOR SUPERIA 200, AGFA VISTA on colour means C-41 and ICE-safe. ILFORD HP5 PLUS, KODAK TRI-X PAN, AGFA APX, VERICHROME PAN on B&W means silver-image and ICE-unsafe.
3. Inspect for mould bloom under raking light. Take a strip to a window and tilt it so light catches the emulsion side at a low angle. Mould shows as star-shaped or fern-shaped white bloom that is not on the film base side. If you see any bloom or smell any musty odour, flag the affected negatives for the lab.
4. Look for any yellow Kodak slide mailers in the same box. If you find them, the slides inside are likely Kodachrome (K-14) and need the Kodachrome-specific scanning path — never ICE-on.

With that triage done, the brief to whichever UK lab you choose becomes specific and unambiguous: "box contains ~120 C-41 colour negatives that can run with ICE Pro on; ~30 black-and-white silver negatives that must be scanned with ICE off and multi-sample 16×; one strip of suspected Kodachrome slides; please flag any negatives where you see mould bloom before scanning rather than running ICE over them." That brief, written exactly that way, will get you a different (better) outcome than a generic "please digitise these family negatives" at every UK lab including ours.

How the EachMoment lab actually handles a damaged-negatives box

For full transparency on what a Coolscan 9000 ED ICE pass looks like in practice when the box is heterogeneous — which is the real-world case — here is the workflow the lab runs. Every Memory Box that comes in with negatives is logged in by a technician who applies the four-step triage above before any film touches a scanner.

1. Triage and sort. One technician with a loupe and raking light separates the box into colour C-41 / colour E-6 / silver B&W / suspected Kodachrome / damaged piles. Damaged includes mould bloom, deep through-emulsion scratches, water-spotted, and adhesive-bonded (where pages of a Boots wallet have stuck to the emulsion).
2. Wet-clean the damaged pile. Anything in the damaged pile gets a PEC-12 + cotton swab cleanup under microscope inspection before it goes anywhere near a scanner. This is the step ICE cannot do for you and the step most "from 17p" UK services skip.
3. Scan colour with ICE Pro on, multi-sample 4×. Kodak Gold, Fujicolor, Agfa Vista, Ektachrome, Provia, chromogenic B&W — all on the Coolscan 9000 ED with ICE Pro four-channel, multi-sample 4×, 4,000 DPI on 35mm or 3,900 measured on medium-format.
4. Scan silver B&W and Kodachrome with ICE off, multi-sample 16×. Different scan pass on the same machine. Multi-sample 16× compensates for the noise floor that would otherwise be reduced by ICE; the result is the cleanest possible silver-image scan with the original grain structure intact.
5. Manual cleanup pass in Photoshop. Two operators review every negative in the silver-B&W and Kodachrome categories for residual dust/scratches and clean by hand using the clone-stamp tool. This is the slow step but it is the only honest way to recover silver-image film at archival quality.
6. Deliver as 16-bit TIFF master + JPEG reading copies. The TIFF is the archival master; the JPEG is what the customer actually looks at on the supplied USB stick. Both are provided.

If you would prefer to send the work in: pricing starts at £0.89 per frame on a standard family roll dropping to £0.53 per frame at the largest volume tier; both passes (ICE-on for colour, ICE-off + 16× for silver) are included at no extra charge. You can request a negatives quote here or order a Memory Box to ship the work to the lab.

Frequently asked questions

The summary in one paragraph

Digital ICE on the Nikon Super Coolscan 9000 ED is genuinely good at what it does on the film stocks for which it was designed — dye-coupler colour C-41 and E-6 negatives and slides, plus chromogenic black-and-white. On those films it removes 85-93% of emulsion-side scratches and dust with no visible loss of image data, which is why every UK competitor mentions it on their pricing page. But on silver-image black-and-white film (Tri-X, HP5, Verichrome Pan), on Kodachrome slides, and on mould-damaged emulsion, ICE produces a worse output than ICE-off + multi-sample 16× + a manual cleanup pass. A UK lab that scans your box with a single ICE-on pass over everything is doing the wrong thing on ~30% of the frames in a typical mixed archive. The brief to give any lab — including ours — is the four-step triage in this article: sort colour vs B&W, read the edge stamp, check for mould, look for Kodachrome mailers. Then ICE is a tool, not a magic word.