Large format art print – but how?

scrabble resolutions

As part of Felix Pruner's NFT campaign, which we are allowed to oversee, the artist approached us with the following request:

If a collector purchases all 8 NFTs from the collection, that collector shall receive a large format print of a work.

Specifically, this is the work:

in the original 60 cm x 50 cm

For the art print, however, the artist wanted a size of 180 cm x 150 cm.

Of course, for such an oversize you need correspondingly high-resolution photos.

The existing photo of the work has a net of 15 megapixels - absolutely sufficient for normal purposes - but for our oversize this only results in approx. 60 DPI. For quality reasons, however, we want to achieve at least 200 DPI.

Here we exceed the limits of what conventional professional cameras can achieve: we need around 200 net megapixels.

Of course, the image must be free of the artefacts and blurring that conventional enlargement techniques would cause:

This quality is forbidden for a high-quality print

AI to the rescue

So: Cameras and conventional techniques fail.

In the field of artificial intelligence, there are solutions under the keyword "Super Resolution".

So what we did is – to put it simply – the following:

We trained an artificial intelligence with a high-resolution Pruner work (15-20 megapixels) and a 3x smaller copy, so to speak, backwards.

So that the AI then knew: “if the small picture looks like this – then the big picture must look like this”

We then gave the AI the high-resolution image as input and the AI then created an enlarged copy of it.

The difference in quality compared to the "conventional method" is clear

We have now received an image that is approximately 16,000 x 14,000 pixels in size, which corresponds to approximately 230 DPI for our target dimensions of 180 cm x 150 cm.

All without any noticeable artefacts or quality limitations.

Leave a Reply

Your email address will not be published. Required fields are marked *