19 August 2012

Why You Can’t Make Digital Look Like Velvia 50

OK – I think I have the best test case yet to demonstrate why you can’t make a digital file look like Velvia 50 (at least with the Canon 5Dmk2 I used for this shot). Take a look at the following image. On the left is the Velvia 50 and on the right is the Canon 5Dmk2. Now look at the colour of the fern and then the colour of the lichen on the tree. When we colour correct the digital file I think you’ll agree it’s impossible to make the lichen and the fern look like Velvia.

This was all done while working on an article on film simulations but I should add that this isn’t to suggest that it’s a good goal to try to make digital to look like film. However, it is dissapointing that some digital cameras can’t differentiate different types of green, especially in vegetation. However, some cameras are better than others and in a future issue of On Landscape we’ll be comparing the Canon, Nikon and Sony (and hopefully a few others) to see which do the best job.

Comments (skip to bottom)

14 Responses to “Why You Can’t Make Digital Look Like Velvia 50”

  1. On August 19, 2012 at 6:06 pm Jonathan Perkins responded with... #

    As usual, interesting investigation Tim.

    I’m not surprised there’s some difficulty in resolving some subtle hues – whilst I don’t have any hard information, my years in telecoms has taught me that building ideal bandpass filters is not possible. The RGB Bayer filters cannot be perfect ‘brick-wall’ bandpass filters with flat passband and vertical sides.

    So taking the green filter for example, and illuminating it with 2 different close shades of green, those hues will be represented by RGB vectors of mainly green and small amounts of red and blue. But level of green component measured will depend on the passband shape of the green filter. That filter response will then distort the resultant RGB vector – if you are unlucky the distortions across the 3 filters could end up with 2 hues mapping to the same RGB point.

    Since I’ve nothing to do with camera design this is of course all speculation on my part, but I’m quite happy to believe that non-ideal colour filters in front of the sensor will be detrimental to the colour fidelity of the camera 😉

  2. On August 20, 2012 at 9:43 am Tim responded with... #


    I’ve followed this thread with some interest, including having a shot at your .CR2 file and not coming within a million miles of the velvia result myself – especially paying attention to the orange/yellow/greens in the relevant bit of grass. But I now have some thoughts:

    It’s one thing to say digital emulations fall short of the real thing; it’s quite another to say it can’t be done. There’s an analogy to be had with audiophilia – there’s a fantasy that says “you can’t get vinyl results on CD”, any number of double-blind tests that can be performed and still the mass of the market prefers to cut from the vibrating strings, via purification in a studio, straight to “quite good enough for my needs” digital recording, bypassing artificial scratches and warping of the media. The correspondence with this Velvia thread should be apparent – two recording media, one analogue one digital, some notion of “realism”, some notion of “good enough for the masses”, and heck if I wanted to emulate the full LF Velvia experience I’d duplicate the layer in PS and flood-fill in black (forgot to remove darkslide) and superimpose multiple disparate images on top of each other for authenticity, too.

    But ultimately both film and digital are made things, with only a finite number of differences between them – lens, sensor (microlenses, bayer, gamma/characteristic curves in as many channels as you care to name, bit-depth in both L and a+b) – but the great unifying factor is, quantum physics aside, in a decent test it should be *the same light* coming from the scene in both cases, so the difference can eventually be pinpointed one way or another – and the magic that can be named is not truly magic.

    I think it’s worth an experiment – a small matter of LF gear being packed away for moving house atm :/

    • On August 20, 2012 at 10:23 am timparkin responded with... #

      Thanks for the comment – I’m not sure what you mean by ‘in a decent test it should be the same light’ – The light here was completely flat overcast and the photos were taken within minutes of each other.

      Also there is a big difference between sound and image. Colour is a perceived sensation that is made up with a spectrum of light is detected by a group of different spread wavelength detectors. This throws away huge amounts of data about the scene. The the sensors (eye/camera) don’t use the same type of detectors, the conversion cannot be the same but only similar – to some extent of similar.

      I’d love to publish your conversion of this image if you have the time – here’s the cr2.

      I’d really like

      • On August 20, 2012 at 11:44 am Tim responded with... #

        `The same light’ – well, a good test will have identical light from the scene at every location, with no fluctuation of any kind. The search for precision is limited by the inability to recycle the same photons for both devices – a 15s gap between exposures in the landscape is a huge margin for error.

        One’s eyes react the same whether the chosen source is film or digital, or the medium is print or screen. They don’t enter into the matter.

        I suspect my analogy was unclear, sorry. Audio is absolutely made up of a spectrum of sound detected by wavelength detectors, and both images and sound are filtered through one’s perception. But rather, my point was the search for magic due to the medium that probably doesn’t exist. That exists in both the photographic and audiophile worlds.

        I’m afraid my attempts at that image were so far off I deleted it outright. However, I will return to it when I’ve done a few experiments. :)

  3. On January 5, 2013 at 7:14 pm Gary Mercer responded with... #

    Hi. There is one camera that seems to really do a nice job differentiating nicely between the greens and lends itself to film simulation—especially Velvia. Thats images from the DP2M and Sigma SD1M.

    I agree that film has its charms—now if I could only find a lab to do it and not destroy the negatives and charge me an arm and a leg. Am I asking too much? LOL

    • On January 5, 2013 at 7:29 pm timparkin responded with... #

      We’ll be testing the Sigma against film and other digital cameras quite soon so keep an eye out!

  4. On February 7, 2013 at 3:36 pm Mikael Otvala responded with... #

    There is one basic flaw it this comparison: On the left IS NOT Velvia 50, is is a digital scan of Velvia 50.

    What colors are lost and what are distorted in the process is anybody’s guess.

    • On February 7, 2013 at 4:10 pm timparkin responded with... #

      It’s a drum scan using Howtek drum scanner and a Hutch calibration target and Profilmaker and then checked on a Just Normlicht daylight corrected lightbox and a against an Eizo monitor. It’s as close as you’ll get without sitting in my office.. I haven’t seen any exhibitions that show the original transparencies so the ‘original’ colours of the transparency are irrelevant to all intents and purposes.

  5. On February 8, 2013 at 8:22 am Mikael Otvala responded with... #

    As the idea of the whole blog posting is to compare Velvia 50 to (in this case) Canon 5D2, saying “the ‘original’ colours of the transparency are irrelevant to all intents and purposes” is quite surprising. We are not comparing film to digital, we are comparing digital to digital, one of them with extra analog (film) stage between the original and the end result. No surprise they are not identical, and that they can not be made identical by tweaking just one of them. Is there a real need to make them identical is my final question.

    • On February 8, 2013 at 8:58 am timparkin responded with... #

      It depends how much you like the velvia or dislike the Canon results. The goal might not be to make them look the same but to emulate some of the characteristics of one of them. i.e. Velvia seperates tones of green very well giving tonal interest in woodland or grassy scenes. It’s almost impossible to emulate that seperation if those greens start off almost the same.

  6. On December 11, 2013 at 2:14 am Chris Raymond responded with... #

    found this page after commenting on the original comparison Velvia vs Canon 5D.

    I understand about the difference in the rendering of the greens and have a possible cause.

    Could it be that film (Velvia) reacts to light frequencies beyond what the eye can see (UV/IR), causing changes in green colouration that is not detected by digital sensors due to the strong IR filters used in digital cameras.

    The clue that lead me to this theory is that using digital sensors in the IR spectrum (Hoya R72, etc) has shown that foliage is very reflective to IR and quite variable in its reflectance – hence if film was reacting to this variation it may cause different foliage types to hue shift by different degrees……the lichen changing by less than the ferns.

    Anyway, not sure if anyone else has raised this idea.

    I’m glad I found your site, plenty of good images and interesting discussion on all things photographic.


    • On January 27, 2014 at 8:34 am timparkin responded with... #

      Agreed.. It’s definitely that sort of change. It might be metamerism within the visible spectrum though as I see it even with a sharp ir/uv cut filter.

  7. On January 27, 2014 at 4:40 am Nick Hawkes responded with... #

    I was able to come rather close to making the Canon version match the Velvia version by adjusting two layers in the Canon image – one to match the lichen “yellow” and one to match the fern “green”. A gradient mask was then added and the two layers were blended normally. The gradient mask was the key, which leads me to believe that the difference between the two shots was mostly due to a change in light between the shots, which caused the coloration of the upper (sunlit) half of the scene to change as the position of the sun changed. Note in this regard the distinct change in the lit side of the tree trunk from a yellow to red.

    • On January 27, 2014 at 8:35 am timparkin responded with... #

      Yes the problem is mostly the greens but the images were only about 2 minutes apart ( enough time to remove ebony and put 5D on) so the difference could well be how the camera reacts to uv/ir

Add your response