1. Reminder: Please use our affiliate links for holiday shopping!

What is "micro contrast"?

Discussion in 'Open Discussion' started by WT21, Apr 4, 2013.

  1. WT21

    WT21 Mu-43 Hall of Famer

    Feb 19, 2010
    Boston
    I've been thinking about what this fabeled "micro contrast" is.

    I feel like I know it when I see it, but I wonder why we don't have a measurement or metric. Why we can't say "this lens' micro contrast is 95" or "this has a 78% micro contrast" -- but I wouldn't even know what it measures!

    I know some have said it's bundled into the MTF numbers, but that doesn't really help me, as I've also heard some folks say you can have a sharp lens, but not a contrasty lens (which has been my experience generally with Oly lenses).

    Well, I still don't have an answer, but would love to hear any input, especially as it relates to m43 gear. Here are some helpful points I found:

    From Lenses with best micro contrast - FM Forums post #7 has "It's high MTF at low frequencies. Which means that a bright area doesn't "infect" a dark adjacent area with its brightness. This isn't necessarily connected to high resolution (which is high MTF at higher frequencies). "

    From terminology - What is Micro Contrast? - Photography Stack Exchange
    "Micro-contrast refers to contrast as measured between adjacent or nearly adjacent pixels. It is often perceived as sharpness.

    Contrast usually refers to the contrast of the entire image which is an indicator of the captured dynamic-range."

    This is also interesting, though not directly related

    Understanding Lens Contrast

    The issue is further confused by color, since color sometimes functions similarly to contrast. Imagine two areas in an image of similar value, but one red, and one green. Take a picture of this with black-and-white film, and you have one undifferentiated gray. Take a picture of it in color, and the green area easily stands apart from the red area and vice-versa. Although it has nothing to do with optical or sensitometric contrast, color contrast helps with definition and hence with a sense of general image clarity. What this means is that different lenses perform differently or perhaps I should say "to different tastes" in black-and-white and color. I conjecture that Leica designers used to pay most attention to relatively low-cycle contrast, meaning in the 5-20 lp/mm range, and then let resolution fall where it may. This is the smartest approach (in my opinion) for black-and-white film. Lenses which have been optimized this way look best for black-and-white. But now that so many people are shooting in color, giving a little more weighting to resolution at higher frequencies (as, say, Canon and Mamiya seem to do pretty consistently) and expecting color to "help with contrast" is a smart approach, too.

    and

    You can have a lens of very low contrast that can be made to transmit the same overall range of light to dark or white to black as one with high contrast. It will just show much less micro detail in the scene, and look relatively muddy and lifeless. Some pictorialist-era pictures actually have a full range of tones from white to black but show (by design) exceptionally low degrees of what we would call lens contrast. Low lens contrast is also created when you put a "softening" filter on a lens you can still print the picture with an overall contrast from pure white to maximum black, but the microcontrast will be severely curtailed.

    So, contrast or dynamic range is about the tonalities in the image, but lens contrast which is the same as micro contrast (??) has to do with the lens' ability to cleanly resolve different tonality changes between pixels. If the lens (and sensor combo??) allows, for instance, lighter pixels to "creep" over into neighboring pixels (perhaps ones that should be darker) than it could be said to have lower contrast.

    How is this different from playing with the contrast slider in, say, lightroom? Does the contrast slides just push out the histogram/tone curve so that darker tones head to the left and lighter tones head to the right, thereby reducing the spread in the mid-tone section? Can the contrast slider ever make up for a deficiency in lens contrast, or does it only address the DR and contrast of the image itself? In LR, is the clarity slider the one that best helps address any lack of micro contrast? (in that it is trying to balance sharpness, contrast and a few other items at the same time).

    Last question -- if you want to test a lens for lens contrast, or more specifically test two lenses against each other, is it best done in B&W to avoid the "apparent" contrast confusion that color can bring, or if you are shooting color, is how a lens renders color an important part of the formula for apparent micro contrast?

    Any technicians out there?
     
  2. kwalsh

    kwalsh Mu-43 Top Veteran

    775
    Mar 3, 2012
    Baltimore, MD
    You've got it well summarized I'd say.

    Remember, lens "resolution" is just picking the MTF50 point off of the MTF curve. It says nothing about what happens at lower spatial frequencies. And 50% response is really, really poor from a contrast point of view!

    I would consider three regions of the MTF curve:

    Contrast - The MTF at the lowest spatial frequencies. Imagine a bright screen with a dark box in the center taking up a large fraction of the FoV. Like this:

    IMG_2802-L.

    The ratio of the bright area to the dark area would represent the lens contrast.

    Micro-contrast - Similar test, except now the dark box is much smaller and there are lots of them. Imagine a fine checkerboard pattern. For 16MP sensors these boxes would be tens of pixels on a side. The ratio between the dark and light would be micro-contrast. Of course we haven't defined a spatial frequency to measure at, and the MTF curve will be sloping at this point. If we wanted to come up with a measurement for "micro-contrast" we would want to define a spatial frequency and then report the MTF value at that frequency for each lens under test.

    Resolution - This also defines a point or region on the MTF curve, but it is defined "the other way around". We ask what spatial frequency does MTF drop to 50% - rather than picking a spatial frequency and reporting the MTF value. We could instead define a new term - "nano-contrast" and pick a spatial frequency close to the resolution of most lenses and then report the MTF value. We'd probably get values between 40% and 60% for most lenses if we chose the spatial frequency correctly. This would be measuring a similar region of the MTF curve as "resolution" is but we'd just be holding the other axis fixed on the MTF curve for the measurement.

    I've some old books here on large format lenses that show nice example images shot with lenses that have identical resolution but very different MTF at lower frequencies. As a real world example consider mirror lenses which can actually have quite high resolution but have very poor micro-contrast and contrast. It is the low micro-contrast that makes them appear much softer than an MTF50 number would imply. See the MTF curves in the figure on this page:

    Mirror Lenses - how good? Tamron 500/8 SP vs Canon 500/4.5L

    That should give you an idea on how multiple lenses might have the same value at one point on the MTF curve but be drastically different at lower spatial frequencies.

    As a last bit, realize that there are three PP controls for these regions as well. You have your global contrast setting - that is like compensating for the lowest part of the MTF curve. You have sharpening, which is attacking the highest spatial frequencies. In between you have what LR/ACR calls "Clarity" and other tools other names. These are what are often call "Local Contrast Enhancement" controls. They are like sharpening with a low amount and large radius. This is attacking the middle portion of the MTF.

    Finally, realize that when we look at an image or "perceived sharpness" or sometimes "acuity" is not solely a function of the MTF50 measurement. Our eyes actually don't care much at all about MTF50. It is really the contrast at lower spatial frequencies that give much of the impression of "sharpness". For this reason micro-contrast has a lot to do with whether an image appears sharp or has "pop" to it.
     
  3. Mellow

    Mellow Mu-43 All-Pro

    Aug 27, 2010
    Florida or Idaho
    Tom
    Great discussion, thanks. I've been wondering about this myself.

    One question (and the mirror-lens example made me think of it): is it possible to enhance micro-contrast with post-processing? I can dial up the contrast or "vibrance" (in LR) and dramatically change the apparent contrast in my images. Does this happen at the expense of resolution? I always found with mirror lenses that post-processing was always necessary, and almost always produced an image with pleasing contrast, even though that's not what was captured on the sensor.

    Finally: I'd love to see a series of images demonstrating variations in microcontrast, just like we can see images with good and bad resolution.
     
  4. arad85

    arad85 Mu-43 Veteran

    477
    Aug 16, 2012
    We perceive sharpness as the ability to change from light to dark quickly (i.e. in a small area). A lens that has a poor MTF or poor ability to render extremes of light and dark will be described as soft. A lens that has good microcontrast is generally considered sharp as it can go from dark to light again quickly (in x and y rather than time...).
     
  5. drd1135

    drd1135 Zen Snapshooter

    Mar 17, 2011
    Southwest Virginia
    Steve
    I'll inquire further: how does microcontrast relate to accutance?
     
  6. kwalsh

    kwalsh Mu-43 Top Veteran

    775
    Mar 3, 2012
    Baltimore, MD
    The "Clarity" slider in LR is the best adjustment for micro-contrast. More is not always better, many great "portrait" lenses of the past actually have poor micro-contrast by design (micro-contrast enhances skin blemishes) while still having good resolution. You can create the same effect in LR by using negative clarity and boosting sharpness. If you look at the presets for the local adjustments brush you'll see a "Soften Skin" preset which is negative clarity with a boost in sharpening.

    On the flip side the "3D effect" much discussed by Zeiss shooters has a lot to do with good microcontrast - and I suspect also good micro contrast in out of focus areas. Dialing up "Clarity" will help achieve the same thing in PP.

    Low microcontrast or contrast also affects color. It lowers saturation by "mixing" adjacent colors together. Most contrast and clarity controls are restricted to luminance only effects by design. If you want to get back some of the color "punch" lost to a low micro-contrast lens you may have to tweak the vibrance or clarity sliders as well.

    The costs of hitting these controls isn't really a loss of resolution but an increase in noise. Much the same way sharpening makes noise more apparent. Shoot a poor lens in good light at base ISO and you can hit clarity and sharpening pretty hard to try to recover. Do the same thing at high ISO and when you start tweaking those sliders you'll start to amplify the noise objectionably.
     
  7. dav1dz

    dav1dz Mu-43 Top Veteran

    926
    Nov 6, 2012
    Canada
    Great info here.
     
  8. drewbot

    drewbot Mu-43 Top Veteran

    702
    Oct 21, 2011
    Toronto, ON
    I believe on FM there is actually a good thread about the subtle differences between the Zeiss vs. Leica M's. I will try to find it since it adds to your argument.
     
  9. drewbot

    drewbot Mu-43 Top Veteran

    702
    Oct 21, 2011
    Toronto, ON
    I think I've had the opposite feeling than you.

    Take the P20 vs O17/1.8 for example.

    I feel the P20 is very sharp, but not as contrasty.

    By contrast (pun intended), the O17 feels more contrasty but not as sharp.
     
  10. RobWatson

    RobWatson Mu-43 Hall of Famer

    MTF says it all. MTF is very difficult to understand. Human perception adds another layer of confusion which is where the effects of color and accutance can turn a 'poor' lens (MTF wise) into a 'good' performer as in a portrait lens.

    Accutance in particular causes my problems since the results of boosting accutance certainly degrade MTF but a percieved dramatic boost on sharptrastness is evident (add sharpness+contrast='sharptrastness').

    In such applications fairly removed from photography like machine vision MTF is critical for the ability for software to extract concrete info from an image so the applicability of MTF is more readily apparent.

    For me micro contrast is the MTF near the resolving limit of the sensor (basically pixel to pixel contrast). A lens with modest MTF in this spatial regime may result in limited microcontrast but when mounted to a camera with a sensor with larger pixels have much better microcontrast - not just a property of the lens but of the whole imaging system.

    Ofcourse, you pornography definition is perfactly applicable -you know it when you see it - just don't spend so many hours on the 'net looking for it. You'll go blind.
     
  11. kwalsh

    kwalsh Mu-43 Top Veteran

    775
    Mar 3, 2012
    Baltimore, MD
    I think one should be very careful about defining or even thinking of micro-contrast or any other IQ figure of merit in such ways. It is almost immediately irrelevant because no one views an image at a 1:1 pixel view, people view an image at a particular size and viewing distance. Defining the spatial frequencies of micro-contrast as dependent on sensor resolution leads to the following counter intuitive and misleading assertions:

    - Images from the same lens but taken at different resolutions then printed to the same size that appear identical to a viewer will have different measured levels of micro-contrast. If the viewer's experience is identical but the measurement is different then that definition of the measurement is not useful.

    - Related to the above, a higher resolution camera will have "worse" micro-contrast than a lower resolution camera. Not only is this misleading it is actually entirely wrong. Higher resolution cameras and their AA filters have a better MTF meaning that they actually improve contrast at a given spatial frequency for any lens attached to them.

    As a direct example, your definition would mean that the 25/1.4 would have better micro-contrast on a 12MP G1 than on a 16MP GH3. The reality is that if we viewed an image -at any size and viewing distance - taken by those two cameras the contrast of high spatial frequencies would be higher on the GH3 image. Exactly the opposite of your definition. To me that is a poor definition.

    This is critical point of nearly every lens test on the web. Essentially everything you see on the web is a measurement of a lens/camera system, not the MTF of the lens alone. Good write up here:

    Notes on MTF measurements

    And yes, I agree that rather than obsessing about numbers and definitions "you know it when you see it" is the most practical approach to evaluating lenses. We already have a slew of web sites producing barely useful MTF50 plots and if you read the forums it is clear the lion's share of the readers of such review sites are drawing erroneous conclusions from that data and its presentation.
     
  12. RobWatson

    RobWatson Mu-43 Hall of Famer

    If so then micro contrast is of no concern in such cases.

    My point regarding pixel size was that if the real limiting MTF comes from the lens then more smaller pixels add nothing except pixel count but no improvement in the resulting image - whether printed or pixel peeped. Perhaps I did use a poor choice of words. The micro contrast would be not better in the case of larger pixels but certainly it would not be better in the case of smaller pixels. Since a whole bunch of 'empty' pixels is 'bad' then such a case is 'worse'. Also, these 'empty' pixels may be empty in terms of improved MTF but they do have noise that may alter the perception to the point of apparent reduced micro contrast - noise can be distracting.

    To your example of the PL25 - I was thinking more along the lines of legacy lenses deigned for much larger pixels/film grains. The PL25 designed to function optimally on the16MP will result in less micro contrast on the 12MP since now the sensor is the MTF limiter - not the lens.
     
  13. kwalsh

    kwalsh Mu-43 Top Veteran

    775
    Mar 3, 2012
    Baltimore, MD
    I guess I have trouble reaching that conclusion since the concept of micro-contrast as existed since well before the existence of pixels and digital sensors. And the concept has in the past been applied to film enlargements vastly smaller than 1:1 viewing of a 16MP image. Micro-contrast effects are quite easily seen on even modest enlargements.

    Got it, and agree. But that is only true to a point. Recall that the MTF of any imager, with or without an AA filter, is not a perfectly flat MTF. It has significant roll-off well before the pixel pitch. And that is in the region historically referred to as "micro-contrast". So it is difficult to define that a lens is limiting the MTF. The two MTF's are multiplied together and so the sensor MTF always degrades the lens MTF. Certainly there is a point where the lens MTF is awful enough that we consider the sensor MTF to be insignificant, but we need to be careful when considering what that limit is. It is non-obvious and typically when people online claim a "sensor out resolves the lens" they are mistaken to claim the extra MP are useless. They are in fact still very useful. Zeiss has a nice write up on MTF and the interplay between film and sensors and the lens:

    http://www.zeiss.com/c12567a8003b8b6f/embedtitelintern/cln_30_mtf_en/$file/cln_mtf_kurven_en.pdf
    http://www.zeiss.com/C12567A8003B8B6F/EmbedTitelIntern/CLN_31_MTF_en/$File/CLN_MTF_Kurven_2_en.pdf

    Pay particular attention to the analysis starting on page 22 of the second part.


    Not the case, see above.

    Again, not the case. Increased pixel density does not make image noise worse. It makes pixel noise worse, but noise spectral density worst case stays the same (and this is what would be distracting). Actually in some cases increased pixel density reduces the noise spectral density of read noise and thus actually makes things better. And of course this is also the experience of the past decade, sensors have become denser and denser in pixel density with greatly improved noise performance at the same time. The common claim "the sensor would be better if the pixels were bigger" is erroneous and the reasons for it well documented.

    Again, there is no clear cut limiter to the MTF. The sensor MTF rolls off across a wide range of spatial frequencies. It is certainly true though that most legacy lenses are not going to perform nearly as well as something like the PL 25 and the sensor MTF effects would most probably be a lot less noticeable in the case of a legacy lens (especially when shot wide open which is what most people want legacy lenses for).

    Anyway - probably have digressed quite a fair distance from the original question in the thread! Apologies for wandering further away from the topic.
     
  14. Savas K

    Savas K Mu-43 Top Veteran

    784
    Jan 10, 2013
    This is a thought provoking discussion. I have nothing relevant to add but it's gotten me thinking.