Olympus 17/1.8 review

Started Nov 17, 2012 | Discussions thread
noirdesir
Forum ProPosts: 10,981
Like?
Re: Macrocontrast vs. Microcontrast
In reply to Anders W, Nov 18, 2012

Anders W wrote:

noirdesir wrote:

Ming Thein wrote:

I think there may be some confusion between microcontrast, macrocontrast and resolution here. On the full size images, the 17/1.8 has slightly better microcontrast in the center than the other two, but macrocontrast is definitely lower.

Ming,

A few years back I tried to figure out a definition for the term microcontrast. Since the most comprehensive way to measure contrast of a lens is a full MTF curve, lenses with different amount of microcontrast should have characteristically different MTF curves. To this day, I keep being surprised that I have not seen a single person talk about this (how microcontrast is reflected in a MTF curve).

I showed a generic example of two lenses with very different character in this post:

http://forums.dpreview.com/forums/post/35584548

and discussed it further in this post:

http://forums.dpreview.com/forums/post/37010026

Of note is that all full MTF curves that I have seen, eg, from DxOmark, always start at one, ie, 100% contrast. That is obviously not really correct as every lens + camera system has some internal flare. While one naturally can normalise all MTF curve to one, this would clearly paint 'foggy' lenses in a much better light than it should be. My idea of what you mean with macrocontrast would be the contrast at essentially target resolution of zero, whereas microcontrast would be the contrast at finer resolutions, whether that is the contrast below the resolution of the MTF50 value as I insinuated in the first post I linked to above or the contrast just above that resolution, I don't know.

And since nobody has rigorously defined microcontrast to my knowledge yet, so far I can only answer that by comparing lots of full MTF curves of lenses which are claimed to either have good or bad microcontrast to see whether there is a pattern (or whether the term microcontrast either covers a wide array of lens behaviours or is just not used very consistently).

It is certainly true that the usage of the term "microcontrast" is not well defined. However, it is obvious to me that it must originally have been coined in contradistinction to global contrast or macrocontrast, and thus refers to contrast at frequencies (resolution criteria) some distance away from zero. At the same time, I have the impression that at least some people use it in reference to contrast at fairly low frequencies rather than very high ones, lower frequencies arguably being more important for perceived image quality at ordinary display size.

It is my impression that it has been coined both in contrast (no pun intended) to global contrast AND to resolution. If we take MTF50 as resolution, this would mean the contrast somewhere between 0 and the 50% mark but some people (like DxO or DPreview with their extinction resolution) consider a much lower contrast for their 'resolution' definition.

As I referred to in my first linked post, the difference between Leica and Zeiss has sometimes been referred to as Leica focussing on resolution and Zeiss on microcontrast (hence my attempt to visualize what might lie behind these different labels people put on lens characteristics).

Like you, I have come to the conclusion that we need to consider the entire MTF curve for a better understanding and I have therefore, like you, looked with interest at the curves now regularly published by DxOMark as part of their lens reviews:

My conclusion after having gone through this exercise for quite a few lenses is that good lenses tend to have MTF curves rather similar to the red curve in the hypothetical diagram you link to, i.e., the curve is fairly close to a line sloping downwards at an angle approaching 45 degrees. Bad lenses tend to have curves more resembling a hyperbolic function (1/x), i.e., they decline very rapidly as the resolution criterion increases from zero to some 10 or 20 lp/mm and then taper off more slowly as the resolution criterion increases further. I can't remember seeing a curve resembling the blue graph in your hypothetical diagram. Would you agree with this somewhat schematic description of mine?

Yes, it is also my impression that the better lenses are closer to a straight line and less good ones exhibit a (larger) positive second derivative (except for the very end of the curve). For two lenses with the same MTF50 value, a straight MTF curve clearly has higher contrast than one that is negatively deviates from that straight line. A purely diffraction-limited lens would exhibit a straight line, which is another indicator that getting close to that straight line might indicate very low other aberrations.

What I have seen has left me more comfortable with the MTF-50-percent criterion used by most test sites. While the result at that contrast level (which, as you probably know, is said to be chosen because it correlates well with human perception of image quality although I have seen no description of the research that this contention presumably rests on) won't tell us everything, it is likely to be strongly correlated with the results we would see if the contrast requirement were set a bit higher.

I am less happy with the fact that DxO themselves set the contrast criterion used for their lens scores significantly lower than other sites (they use 20 percent rather than 50), both because the differences between lenses are likely to be less clear-cut in this contrast region and because I think it is less important for perceived image quality how a lens does in this contrast region at anything resembling normal display size and viewing distance.

Yes, I too find DxO lens scores a bit puzzling as they do not seem to correlate very well with anecdotal evidence (and while anecdotal evidence is often somewhat biased, overall it does correlate with what matters: how much humans like the final result).

Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow