Interpolation experiment


Interpolation, upsizing or uprezzing, i.e. increasing the size of a file, used to be a common occurance amongst library / stock photographers. In the early days of digital photography, not only was it an accepted practice but one that was positively encouraged by the libraries, so that digitally captured files could be available at the same size as film scans. There used to be lots of programmes that claimed to do this well, though they were all almost exactly the same and no better than just increasing the file size in Photoshop. As pixel sizes for sensors have increased, there seems to be much less of this, but there are vast numbers of files sitting on websites that are the result of upsizing.

It occurred to me a few weeks ago to see whether upsizing files from my current cameras affected their acceptance rate. The reason I usually buy some expensive, large heavy camera is to get a large file size to upload to the websites that sell my work. This is a hangover from my film days, when the MF transparencies I shot sold better, and got used for covers, A3 spreads, Ad. campaigns etc, rather than the 35mm I shot. Libraries in those days were always much more inclined to accept the larger film size since they looked so much better on lightboxes, which was how clients looked at pictures, and the repro houses preferred them for printing at large sizes.

I also wanted to see what reaction I got from the libraries who decide whether to accept or decline the files I upload, when they have no knowledge of what camera was used to take the picture, and have to make a judgement on the quality in front of them. When I produce images created from stitching several files together, I remove the exif data from the files (achieved by using save for web in Photoshop) so that there is no "misunderstanding" about the file size. Libraries these days usually make it clear that they won't accept interpolated files. However, that doesn't take into account the quality of the original image. Files from some cameras upsize very well and after doing so can look better than those shot on sensors with higher native resolution. For example my upsized images from my non-AA filtered Leica M9, when interpolated to the same size as those I shot on my Nikon D3X, looked as sharp, if not sharper.



So what have I been doing? Well I've been upsizing m43 images (both 12 & 16MP) up to 24MP. NEX-7 and Nikon D600 24MP files up to 36MP, and Sigma DP2 Merrill 14MP files up to 24MP or 36MP. In order for this to work successfully, I have been adding no sharpening whatsoever. This results in a "clean", slightly soft file, with no artefacts, which again is what stock libraries like. The theory goes that files on library sites available for sale and download should be unsharpened, since the amount of sharpening required can only be determined when the file is sized for publication. Its a good theory, but in practice digital files are often published unsharpened as I discovered when I ran my own library. When all pictures were scanned and prepared by a repro house, they did their own sharpening, but these days it often just doesn't happen. However, that is the general practice, so thats what stock photographers do (or rather don't do!)

So the files that I have been uploading are somewhat softer than I would like, but the result has been that I have had a higher % accepted than when I upload files at their native resolution. So why might this be?

Ideas about photographic practice tend to stick around whether they have any relevance any more or not. I am sure that those who assess files for libraries are still told to reject files that appear to be "over sharpened" though that is often a matter of taste and opinion. However it is true that the flatter, cleaner and closer to some kind of neutrality a file is, the more useful it is regarded for libraries. This "neutral" file is much more useful to a designer, who can then make any adjustments in Photoshop that they desire. An over-processed file removes a lot of options and consequently is regarded as less "flexible". 

Another factor that influences this is that the vast majority of files sold by picture / stock libraries are less than A4. At least 50% go to web and electronic publication these days anyway. So, if a file looks slightly soft at its largest size, it does "sharpen up" as the size reduces, so that its probably at its optimum at the most popular sizes, whereas a file that has had a lot of sharpening applied can appear over sharpened and often quite unpleasant at these smaller sizes.

Thirdly, there is a difference between how files look on a computer monitor and how they reproduce in print. We all pixel peep files at 100%, but I've been told on numerous occasions by those who work in printing that if you want to check how a file will reproduce in a magazine or a book, view it at 50%. and I have certainly found that to be true.

So a file that looks a bit soft at 100% is not necessarily something that a library will reject, for the above reasons. Add in the fact that digital cameras are now capable of much better results than they used to be and interpolation and upsizing still works. However, it does still require the original file to be good enough to do this with. If anyone thinks they can upsize their iPhone images, then I advise them to think again, unless they are intending to achieve that "hipstamatic, low-fi look" i.e. Complete c**p!

This does mean that I may not have to be looking for the highest MP count any more, which has interesting implications. The Sigma DP2M is obviously a special case and the images it produces punch way "above their weight" but its interesting how well m4/3 images upsize. In a previous post I showed that an upsized OM-D file can look as good as a native NEX-7 file, and again its the quality of the original, the quality of the lenses, the strength of the AA filter etc. that can be just as important as the no. of pixels. 

So next time you are thinking "shall I upgrade from 12 to 16MP?" or something similar, you may not need to. Just upsize the files from your current camera and see how that looks. They may be better than you think.




N.B. to see more on the cameras and lenses featured in this post click on the relevant labels (tags and keywords) below.
All original material on this blog is © Soundimageplus
For comment and discussion - join me over at Google+
about soundimageplus - soundimageplus website
soundimageplus on flickr - http://www.flickr.com/photos/45203414@N06/
soundimageplus blog readers pictures group - http://www.flickr.com/groups/1705334@N24/
soundimageplus on YouTube - http://www.youtube.com/user/soundimageplus
soundimageplus on Vimeo - http://vimeo.com/user1050904/video