Digitising and Upscaling a 40 mm × 28 mm Stamp for Large Format Printing

Fine-art and archival scanning often focuses on preserving small objects such as photographs, negatives or printed artefacts. A common question is how far these objects can be enlarged while still retaining convincing image detail.

To explore this, we carried out a technical test using a postage stamp measuring 40 mm × 28 mm, digitising it and attempting to scale it far beyond conventional limits.

The objective was simple: scan a small printed object and prepare it for convincing large-format printing, initially targeting A2 or A1 size.

---

Digitising the Original Stamp

The stamp was digitised using a high-quality flatbed scanner at its maximum optical resolution of 800 ppi.

Optical resolution is important here. Many scanners advertise higher “interpolated” resolutions, but these simply invent pixels mathematically rather than capturing real image information.

Scanning at 800 ppi optical resolution ensures that the scanner captures the maximum amount of real detail available in the original object.

Given the stamp dimensions, the resulting scan produced approximately:

The native scan therefore produced an image of roughly 1,260 × 880 pixels.

This resolution is perfectly adequate for reproduction close to the original size, but far below what is required for large-format printing.

---

Correcting Optical Blur

All digitisation processes introduce a small amount of optical softness due to the scanner lens and sensor system.

A common method of correcting this is Unsharp Masking, but this approach produces a visible halo around edges. When images are enlarged significantly, these halos become highly visible and degrade image quality.

Instead, we used a detail reconstruction approach based on AI-driven sharpening models. This method enhances micro-contrast and restores edge definition without introducing halos or artificial edge artefacts.

Correcting optical blur at this stage is important because any imperfections will also be magnified during enlargement.

---

Preparing the Image for Enlargement

For practical printing workflows the image resolution was standardised to 300 ppi, the typical resolution used for high-quality photographic printing.

At this resolution the scanned image measured approximately:

120 mm × 84 mm

This is still far below the target sizes of A2 or A1, meaning the image would require significant enlargement.

To reach A1 size, the image would require roughly a 10× linear upscale.

Traditional enlargement methods can technically achieve this, but the results are rarely convincing. Standard interpolation methods such as bicubic scaling simply stretch the original pixel data, producing soft edges and blurred textures.

---

AI-Driven Upscaling

To overcome the limitations of traditional interpolation, the image was processed using a series of AI-based image enhancement models.

These tools are designed to reconstruct plausible detail during enlargement by analysing patterns within the image rather than simply stretching pixels.

Importantly, the workflow did not rely on a single upscale step. Instead we used an iterative enlargement process where the image is gradually increased in size while allowing the model to redefine fine detail at each stage.

This technique helps maintain edge clarity and prevents the plastic or artificial look that can occur when images are scaled aggressively in a single step.

---

Pushing the Limits

Once the image successfully reached A1 size, we extended the experiment further to see how far the image could be enlarged before file size and system limitations became a constraint.

Image formats impose various maximum dimension limits:

Although some formats support extremely large dimensions, practical limits are imposed by workstation memory and processing performance.

For reliable production workflows we generally treat 50,000 pixels on the longest edge as a comfortable working limit for TIFF files.

---

Final Image Size

Using the iterative AI workflow, the stamp image was enlarged to approximately:

3 metres × 5 metres

At a typical large-format print resolution this corresponds to roughly:

35,000 × 59,000 pixels

This is close to the practical TIFF working limit and represents a substantial increase over the original scan.

---

Enlargement Factor

The original artwork measured 40 mm × 28 mm.

The final experimental print size reached 5 metres on the longest dimension.

This represents approximately:

In other words, the printed image is more than 15,000 times larger in surface area than the original stamp.

---

Print Results

Test prints at this scale produced surprisingly convincing results.

Instead of the soft, blurred appearance expected from traditional enlargement methods, the prints show:

• well-defined edges

• recognisable textures

• clearly separated colour areas

• sharp graphic shapes

While the image does not contain the same microscopic detail as the original stamp when viewed extremely close, the overall visual impression remains crisp and believable at normal viewing distances.

This is particularly important for large-format display prints where viewers typically stand further away from the artwork.

---

Practical Applications

Although this experiment was intentionally extreme, the workflow demonstrates useful possibilities for digitisation and large-format reproduction.

Potential applications include:

• Fine art reproduction

• Museum digitisation projects

• Historic document enlargement

• Archival printing

• Exhibition graphics

• Large-scale educational displays

AI-assisted enlargement does not replace high-resolution photography or scanning, but it can significantly extend the usable size range of existing digital material.

---

Conclusion

A 40 mm × 28 mm postage stamp is a very small object. Traditional image processing methods would struggle to enlarge it convincingly beyond A1 size.

Using a carefully controlled workflow combining:

• high-resolution scanning

• optical blur correction

• iterative AI upscaling

we were able to produce a printable image measuring approximately 3 m × 5 m, while still maintaining clear and recognisable detail.

This experiment highlights how modern digitisation and AI image processing techniques can push the boundaries of what is possible when preparing small originals for large-format printing.