3d pan white

Film vs. Digital: An Analysis of Chromatic Fringes

To read this article: view it in full size.


Automatic translations: 繁體中文, Français, Deutsch, 日本語, 한국말, Español


Long story short: The 85 1.2L II was shot wide open in broad day light (to really push it) on a film SLR and a 5D Mk II DSLR to see which medium created images with less purple fringes. Film clearly wins in this sense. The sudden appearance of fringes in overexposed areas (in digital) and not in other areas also suggest this is not lateral chromatic aberration (e.g. see white text on flag vs. SUV)


All the text from the article (image) for accessibility and online language translators:



First, this test does not attempt to compare resolution or the exposure latitude of film and digital. The interest is purely in understanding the appearance (scale) and behaviour of colour fringes (a.k.a. purple fringes) that plague fast lenses, noticeably on digital cameras. Therefore scanning the images any better to improve resolution, shadows, etc. will not help us understand this behaviour any better.


Both the film and the digital versions of the scene were captured few seconds apart. Here are the details:


Cameras: Canon EOS 3 (+ Epson V700 & custom calibrated holders), Canon 5D Mark II


Resolution: 21MP for digital, film scanned @ 54MP then scaled to 21MP. No sharpening applied to either image. 100% crops shown.


Lens: Canon EF 85mm f/1.2L II @ f/1.2 (And no focus changes)


Film/Digital contrast profile: Fuji Velvia RVP 100F, 'Standard' DPP on raw


Sensitivity: ISO100


Shutter speed: 1/3200s


Support: Gitzo GT2541 + Manfrotto 405


Results and Analysis (left: film, right: digital)

Crops 1: The differences here are minimal. It is not certain whether this is because the crops are from the exact center of the image circle (r=0%). Some purple fringing is visible in both.


Crops 2: This set clearly shows the purple "ghost" of the street light in the digital image. The film image barely shows any fringing and its image quality is obviously very high.


Crops 3: These crops are from essentially the same position (r=36% vs. r=35%) on the image circle as the previous (street light) set of crops. However, both shots show some purple fringing on the open car door's glass. The glass here is obviously more reflective than the street light's metal and this could be the reason. So, looking at only these two images, we can see that the film image is not completely fringe-proof but it is more resistant than the digital sensor.


Crops 4: In this set, we hardly see any differences. This was also the case in most parts of the image. I.e. where there were no blown highlights against a dark background, the image quality differences were minimal.


Crops 5: This set shows some odd behaviour. Here, we see no purple fringing on the digital image where we really expect to see some. The pole is dark enough to show the fringes but it does not. Only difference from the previous sets is, here the dark bit (the pole) is in front of the blown highlight.


Crops 6: Another example of the digital image falling apart with fringes. The problem does exist as you go towards the outside of the image circle. Also we don't see the fringes getting on to the dark/black car on the side that is closer to the camera. This is consistent with what we saw in the previous set of crops (with the pole).



Without the need for much explanation, it is obvious that film is much less susceptible to colour/purple fringing than a typical digital camera.


This brings us to the next question: Is this caused by the lens or the imaging medium? From what we have seen, it is easy to think it is the medium that is responsible. But it is difficult to conclude this with full confidence as we have also seen fringes in the film image (crop set 3) but in much lower magnitudes. So from the analysis we have done here, all we can say is that it is possibly a combination of both (the lens and the imaging medium).


As a side note, it is worth mentioning that fringes, aberrations and artifacts cannot be "corrected" although modern camera and popular image editing software claim to do so. Simply put, what is lost is lost. You cannot create information that was never captured. So these "lens correction" software merely hide the appearance of these nasties and in the process, even further resolution and hue losses are possible. Same goes for sensor noise and noise reduction.



Please take a moment to look at my other articles as well.

17 faves
Taken on September 15, 2011