Is lens contrast a tradeoff vs. resolution?

<p>Hi Sarah - I think you are on exactly the right track. Specifically, I suspect that the statement was taken out of a context which stated a restricted domain, a set of conditions, etc. in which it was correct. However, without these, all one is left with is a very general statement, which is clearly incorrect as I demonstrated by giving examples of ways to separately modify the contrast and resolution of a lens with no interaction between these variables. This is why, in my very 1st post in this thread, I offered possible alternative statements that would be correct. At this point, IMHO, the possibility you just offered sounds even more likely than the two I offered in my 1st post.</p>

<p>Tom M</p>

<p><strong>Tom</strong>, I actually have not found in the literature something clearly clarifying. There isn`t a hint about the "trade-off".</p>

<p>What you mentioned above; the closest thing I found is the higher the spatial frecuency the lower contrast; also, the better correction the higher contrast and resolution. There are lenses with high contrast/low resolution, low contrast/high resolution, and with both high (or low) contrast and resolution, depending on the quality and design (aberration correction). I`m now confused; the trade-off issue not that clear.</p>

<p>And that`s right, the MTF charts only goes from zero to one... Thanks.</p>

<p>Tom, we pick up information and read stuff all the time. I've been reading about optics for three decades. I don't storehouse literature in that field. If you think it's just opinion that's fine. It takes enormous amounts of non-productive time to keep an appendix and cite exact sources. I'm sure if you took the time to read Erwin's book instead of selfishly demanding an immediate data dump you would find the material you would like. Again, the material is all out there. Now it's your turn to dig for it.<br>

You come across like a disgruntled debate student and that's never fun for anyone else.</p>

<p>Hi Kirk - My PhD was in optics and spectroscopy. In both a large national lab and in the private sector, I have managed research groups of PhD physicists and PhD engineers since the late 1970's developing and using lasers and optical systems. I have 17 patents in this area in the area of new lasers, laser-based systems, and applications of optics. I have successfully developed many very complicated and very expensive optical systems based on my knowledge of this field, and served for five years on the patent review board for a major national laboratory in the area of optics and laser applications. Eight years ago, I left the private sector and accepted an appointment as a full professor at a major university. I assure you that I have already read quite a bit of very authoritative material on optics and lens ;-) .</p>

<p>With respect to demanding accurate and specific citations, you mentioned that you are active in the area of LEDs. Suppose someone made a claim to you about LEDs that sounded wrong or incomplete. Then, when pressed, they cited a lengthy general Wikipedia article on solid-state devices instead of providing a direct and appropriate reference to an IEEE journal article on LEDs that speaks directly to the issue at hand. In actuality, even worse than citing a general Wikipedia article (which, at least is somewhat peer reviewed), they only cited a bunch of lengthy, un-referreed blog entries all by the same person, and to make matters worse, never once gave a direct quote even from *that* source which backs up their claim. I think you would be telling them in no uncertain terms that they needed to be a bit more rigorous when speaking to you on the topic of LEDs. That is exactly what has happened here, except the positions are reversed.</p>

<p>Cheers,</p>

<p>Tom M</p>

<p>Wow! A lot of words on a subject about visual judgement on lens contrast vs resolution and I still have no visual proof or evidence showing me this exists much less a clear understanding what the point is in this thread.</p>

<p>How about some 100% crops showing the differences between contrast vs resolution between two different lenses. I'm not clear what is meant by lens resolution. I examined in ACR the Zeiss 85mm Raw samples downloaded from the link Kirk provided and I don't see anything that shows this. The $1500 price tag does go for adequate sharpness shooting wide open in shaded light and a decent blurring of the background due to DOF.</p>

<p>It did get me to investigate why those Raw files have a green bias over my Pentax K100D's reddish orange bias in all my Raw shots. I scrolled up to the image of the Zeiss lens in that linked review and sure enough I can detect the magenta color reflected in the glass compared to my Pentax lenses' green. It confirms what I always suspected all along...Trust the sensor electronics, metering and everything upstream from there over what's in front of the sensor. After the A/D converter it's all interpreted anyway including MTF, contrast, sharpness and color.</p>

<p>Kirk, as a fellow Texan living 50 miles south of you I commend you on publishing such straightforward and concise books on lighting and photography, especially the one on LED. You even include image samples. I checked out your books on Amazon. However, low contrast style of image rendering represented in your gallery isn't my cup of tea.</p>

<p>Don't get me wrong, they're very slick professional GQ, Vogue and Texas Monthly quality, it's just viewing them on a 100 cd/m2 calibrated display in dimly lit room on a stark white page makes my eyes adapt to seeing them a bit flat and murky but quite sharp looking. Wonder if you'ld ever considered this the reasons for your misgiving toward the "pop" high contrast look you seem to notice and base your premise on contrast vs resolution with regard to lens characteristics.</p>

<p>Below is a macro shot of baby bird the size of a quarter that had fallen out of its nest I shot with a 1985 Vivitar junk drawer 28-135mm lens attached to a crappy Quantaray teleconverter. As you can see it's quite low contrast, but editing in ACR brings out the definition. Can you tell if that lens system has high resolution and low contrast and vice versa?</p><div></div>

<p><a href="../photodb/user?user_id=313836">Kirk Tuck</a> , Dec 01, 2011; 08:05 p.m. said: </p>

 

<p>Kirk Tuck is absolutely correct. Thank you. Start here:<br /> <a rel="nofollow" href="http://www.luminous-landscape.com/tutorials/understanding-series/understanding-mtf.shtml" target="_blank">(link)</a><br /> Well, this is interesting: <a rel="nofollow" href="http://www.luminous-landscape.com/tutorials/understanding-series/lens-contrast.shtml" target="_blank">(link)</a><br /> As is this: <a rel="nofollow" href="http://www.imx.nl/photo/optics/optics/page63.html" target="_blank">http://www.imx.nl/photo/optics/optics/page63.html</a><br /> This is pretty good: <a rel="nofollow" href="http://www.imx.nl/photo/optics/optics/page93.html" target="_blank">http://www.imx.nl/photo/optics/optics/page93.html</a><br /> <a rel="nofollow" href="http://imx.nl/photo/zeiss/page116/page116.html" target="_blank">http://imx.nl/photo/zeiss/page116/page116.html</a><br /> or this: https://sites.google.com/site/seevve/historical-perspective-on-minolta-lens-design-philosophy<br /> some good tidbits here: <a rel="nofollow" href="http://photo.imx.nl/optics/optics/page81.html" target="_blank">http://photo.imx.nl/optics/optics/page81.html</a><br /> <a rel="nofollow" href="http://www.dyxum.com/columns/articles/lenses/sal-85f14cz/carl-zeiss-sal-85f14cz_review.asp" target="_blank">(link)</a></p>

 

<p>_____________________________________________________________________</p>

<p>I had a look at those links and the ones on MTF do no more than confirm my own understanding of MTF, and the fact that high resolution corresponds to high contrast at small spatial frequencies. Some of the text linked to is very long, and discusses loads of side issues unconnected to your post. I'm afraid I'm not prepared to spend an hour or more reading through text, on what might be a wild goose chase, and I doubt others are. Could you please quote relevant extracts?</p>

<p><a name="00ZgMw"></a><a href="../photodb/user?user_id=313836">Kirk Tuck</a> , Dec 01, 2011; 09:59 p.m. said "Contrast, according to both Leica and Zeiss, comes at a cost of resolution."</p>

<p>_____________________________________________________________________</p>

<p>Could you please provide links or cite your sources for that statement. Especially since it contradicts several of the articles that you link to.</p>

<p><a href="../photodb/user?user_id=1722891">Tim Lookingbill</a> said; "I'm not clear what is meant by lens resolution."</p>

<p>In astronomical circles lens resolution is the ability to separate two point sources. As you move the sources closer together, at some point the images will fuse into one, and they will no longer be resolved. The telescope lens would be used with a high power eyepiece, since the lens out-resolves the human eye. Photographic lenses tend to have a lot of off-axis aberrations, so resolution would depend on the distance from the image centre. In the example images of the chick above, the contrast at boundaries (edges) has been enhanced to make the edges stand out, and give the illusion of greater resolution/sharpness.</p>

<p>Leif, thank you for adding your comments and sentiments on wasting time, which exactly parallel my own feelings on the subject. BTW, I think you made a minor typo ...</p>

<p><em>"...and the fact that high resolution corresponds to high contrast at small spatial frequencies..."</em></p>

<p>I think you probably meant to say <em>"at high spatial frequencies"</em> or <em>"at small length scales"</em>.</p>

<p>Cheers,</p>

<p>Tom M</p>

<p>Tom, thank you for the correction. You are of course correct.</p>

<p>I suspected astronomy would be at the core of all this complexity derived from photographing and viewing objects that don't tell us much as to what we're looking at or is relevant or practical compared to pointing the lens in the opposite direction and that being Earth.</p>

<p>So the only way to test whether a lens is offering high resolution is to see how well a tiny tree renders a mile away in like say a landscape. Is that the gist of it?</p>

<p>Wonder where Kirk went? He's not so chatty all of a sudden.</p>

<p><a href="../photodb/user?user_id=1722891">Tim Lookingbill</a> "So the only way to test whether a lens is offering high resolution is to see how well a tiny tree renders a mile away in like say a landscape. Is that the gist of it?"<br>

In photography the usual way to test resolution is to photograph a lens chart at a well defined distance from the camera. I guess you could photograph a tree, although atmospheric issues enter the equation e.g. air turbulence etc, and you'd have to accurately measure the distance to the tree etc. A lens chart is easier! And before someone mentions it, you are really testing the resolution of the lens sensor pair, unless the sensor easily out resolves the lens.</p>

<p>Wow. Someone looked at my gallery of web images and extrapolated my understanding of resolution versus contrast. Then Tom lectures and corrects everyone. And you wonder why people don't stay and submit to the silliness? Read more, type less.</p>

<p>That's your come back, Kirk?</p>

<p>Here's mine. Write less, show more...</p>

<p>Leif, I think somewhere into this video lecture by the inventor of the CMOS sensor...<br /> http://image-sensors-world.blogspot.com/2011/10/eric-fossums-lecture-on-image-sensors.html<br /> <br /> ...it mentions the exact pixel pair sensor resolution max which I think he says lenses quite out resolve at some megapixel amount. You'll have to pay attention at the point he mentions this.</p>

<p>Kirk Tuck said, in response to Tom Mann: "I'm sure if you took the time to read Erwin's book instead of selfishly demanding an immediate data dump you would find the material you would like. Again, the material is all out there. Now it's your turn to dig for it. You come across like a disgruntled debate student and that's never fun for anyone else."</p>

<p>Kirk Tuck said: "Wow. Someone looked at my gallery of web images and extrapolated my understanding of resolution versus contrast. Then Tom lectures and corrects everyone. And you wonder why people don't stay and submit to the silliness? Read more, type less."</p>

<p>Kirk, the comments by Tom, myself and others on a claim about contrast and resolution that you made in your blog are polite, reasonable and coherent. You respond with obfuscation and childish abuse. You refer to Tom Mann as a 'student', whereas he has the Ph.D. in optics and astronomy, you don't. Although I have a Ph.D. in physics, I am no expert. Perhaps you think that if you make statements in the public domain, you are so important that they are beyond criticism/discussion. Maybe rather than behaving like a petulant child when someone asks a quite reasonable question, you should grow up and address the issue, namely what is the basis for your claim about contrast and resolution. If it is correct, then we have all learnt something. If it is incorrect, perhaps you should have the grace to accept that fact.</p>

<p>If you cannot cope with criticism of writings you put into the public domain, then don't publish them. If you do, then accept that people will pull you up for technical inaccuracies, on the grounds that misleading people is unfair. If you think the criticism is unjustified, then provide evidence for your claims. But having a temper tantrum does noone any favours.</p>

<p>Leif, just to set the record straight, it was optics and molecular spectroscopy (with earthly applications), not optics and astronomy. I have only one short letter to the editor in an astrophysics journal, and that was on isotopic abundances and isotopic fractionation mechanisms as measured in the lab, not real astronomy. :-(</p>

<p>Tim, w.r.t. testing lenses at infinity, this doesn't actually have to be done using objects at infinity. It's often done in the lab by putting a second, *very* expensive and well corrected lens system between the lens under test and a resolution or other target, making it appear to the lens under test that the target is at infinity. </p>

<p>Because the ancillary lens system does not have to zoom, does not have to cover a wide FOV, is not constrained by needing to mate with a particular line of camera bodies, and and because only one of these lenses is needed to test many different lenses, its optical properties are impeccable and any aberrations it does have don't limit the measurements on the lens under test. </p>

<p>Tom M</p>

<p>Eric Fossum's statement about image resolution is correct as far as it goes, but so oversimplified that it borders on meaningless without a lot of qualifications. The statement he makes is that the diffraction limit for green light with an f/2.8 lens is ~4 microns, while cameras now have pixels/sensels down to ~1.1 micron.</p>

<p>That's not <em>nearly</em> as bad a situation as it may initially sound though. In particular, we absolutely do no want the lens to project an airy disc that only hits one sensel (assuming a normal Bayer pattern sensor). Such fine detail would (not might, but would, unavoidably) lead to false color. For example, an Airy disc of white light that size hitting a blue sensor would show up as blue, not white.</p>

<p>At the risk of over-simplifying in another direction, the color of a pixel in the output is based on the value at that sensor and the average of the surrounding sensors of the other two colors (with a bit of manipulation to take into account the fact that there are twice as many green as red or blue sensor wells). That means for the sensor to properly resolve all the information from the lens, the Airy disc has to cover at least three senor wells, and we'd like it to cover a 3x3 array of sensor wells. As it happens, the distance across the diagonal of a 3x3 array of sensor wells with a 1.1 micron pitch is ~4.7 microns. In other words, a 1.1 micron sensor pitch is getting close to the lower limit of size that could produce a resolution increase, but isn't really exceeding the theoretical limit.</p>

<p>The stated assumptions are the killer though. At f/22, even a sensor with much coarser pitch has more resolution than the lens can hope to produce. Going in the other direction, a diffraction-limited f/1.4 lens would resolve (a lot) more than any current sensor.</p>

<p>That assumption is non-trivial too though: I certainly wouldn't expect to see many diffraction-limited f/1.4 lenses any time soon. Though they're generally somewhat slower, I wouldn't expect the lenses used with the finest-pitch sensor (e.g., in cell phones) to be diffraction limited as a rule either.</p>

<p>It's simple. High contrast is high MTF at low spatial frequencies (say 0-20 cycles/mm). High resolution is high MTF at high spatial frequencies.</p>

<p>A perfect lens has both. In fact if it's perfect it has the maximum possible MTF at any given spatial frequency (i.e. all of them). In practice a designer could trade of MTF at different spatial frequencies if he couldn't get a design that maximized both at once without adding more elements and more cost.</p>

<p>High sharpness is a perceived quality involving both high resolution and high contrast,</p>

<p>And still no visuals or comparative image crops to show what is being discussed. Photographers use their eyes not physics diagrams to determine if their lens grabbed all the detail possible.</p>

<p>I figured out from my post processing demo and the many other's I've performed that higher contrast (global, local and micro) means higher perceived resolution=(resolution as defined by the lens directing photons to a set of pixel sites that distinguishes the smallest detail captured). Was that from the lens or my post processing?</p>

<p>Which points to the fact that generally these discussions always point to the lens as the dominant influence without taking into account what demosiacing algorithms (which there are several offering different interpretations in how they render pixel detail) and editing software bring in determining what influences what.</p>

<p>It's comforting and quite exciting to know I'm in discussions with Phd's in Physics and Optics. I feel I'm walking among kings. Thanks for putting up with my obstinate layman style line of questioning.</p>

<p> </p>

<p>Tom, apologies for misrepresenting your thesis subject/area, my excuse is that there is too much in this thread to easily read through.</p>

<p>Tim, I certainly would not describe myself as a king. A Ph.D is a very narrow area of work, and as mine was in quantum theory applied to solids, I have no special knowledge of optics, beyond undergraduate courses, and general reading.</p>

<p>As an aside, I have learnt that contrast can be very important in practical terms. I once owned a Nikon fieldscope (one of the cheaper ones), and although resolution was good on a sunny day, as the light dropped, and became flatter, so the resolution dropped. This is not so surprising, as the eye finds it harder to distinguish detail in a low contrast scene. A better fieldscope, such as one of the high end Nikon ones, or the Leica I own, performs much better in low light, as it preserves more of the contrast present in the scene. The same effect can be seen when comparing budget binoculars with high end ones.</p>

<p>Leif, no apologies needed, I just didn't want people asking me questions on Cosmology, which happens all too often this time of year at Xmas parties when they learn I'm a physicist. ;-)</p>

<p>Leif, w.r.t. Ph.D.s being "kings", I bet you have conquered many a Brillouin zone, but are just too modest to admit it. ;-)</p>

<p>Bob, thanks for checking in. Lief told Kirk Tuck (the subject of this thread) almost exactly the same thing in Lief's 1st posting in this thread. Unfortunately, Kirk ignored it.</p>

<p>Tim, you are definitely on the right track when you suggest that the sensor, any other relevant hardware (eg, anti-aliasing filter), demosaic'ing and other algorithms should all be considered when trying to quantify / compare the performance of an optical system. In fact, this is rather commonly done when quantifying the performance of costly, complex electro-optic imaging systems (eg, FLIRs). Including other system components besides the lens does open up a bit of a can of worms, because it introduces many other variables (eg, what were the values of the parameters in your algorithms), and, unfortunately, ties some of these together (eg, the MTF curve can become a function of light level), that aren't tied together if you just consider the lens, but it's definitely a useful approach. The problem is that most photographers don't even understand MTF curves for lenses, let alone MTF curves for overall systems.</p>

<p>Taking a systems level approach instead of considering a lens in isolation is why, in my very first post in this thread (specifically, the 1st set of material in italics), to try to give Kirk the benefit of the doubt, I suggested that perhaps Kirk's suggested tradeoff was referring to "system", not "lens". In this case, for a given overall system dynamic range and quantization depth, increasing contrast by electronic means would limit tonal resolution in a very tradeoff-ish like way.</p>

<p>Tim, w.r.t. providing example images which illustrate how different lens MTF curves render the same scene, it's not easy to do this because such images must be obtained in a consistent manner, by a reputable / knowledgeable source, must be at the same focal length and aperture, and the visual comparison images should not be confounded by the fact that there are many different types of lens aberrations, and all of these can all lower the MTF curve, but all will look very different by eye. </p>

<p>As an example, look at the photozone.de lens reports on the 35mm f/2 Nikkor AF-D (http://www.photozone.de/nikon_ff/444-nikkor_afd_35_20_ff?start=1), and the correspondng Zeiss Distagon ZF (http://www.photozone.de/nikon_ff/587-zeisszf3520ff?start=1). Scroll to the corresponding sections on MTF. Here we have two lenses of the same FL, same f-number, measured in a consistent manner by a reputable source, but unfortunately, they don't provide plots of MTF vs spatial frequency or vs distance out from the optical axis (at a constant spatial freq). Rather, by a process that they carefully describe (http://www.imatest.com/docs/sfr_mtfplot/), they turn the MTF measurements into a single "resolution" number, and plot this as a function of aperture at three distances out from the optical axis, which is not what you are seeking. In addition, if you look at the cropped images in the Bokeh sections of these two reports, you can easily tell that different types of optical aberrations are at play for each lens.</p>

<p>Hopefully, this one example illustrates the difficulty of obtaining good, reliable data in exactly the form you seek (ie, pairs of example images with certain variables held constant between the two systems under comparison). Hopefully, the above also illustrates that there many important variables that contribute to the overall utility / quality of a lens, and one is treading on thin ice if one tries to make decisions based only on resolution numbers, only on CA measurements, only on crops showing Bokeh, etc.</p>

<p>Cheers,</p>

<p>Tom M</p>

 

<p>Appreciate the responses from both of you on this subject.</p>

<p>I understand your points, but I was trying to get insight from a visual inspection on the connection between contrast and resolution in a lens which is the title of this topic. Just want to know in this regard what to look for in a lens I might purchase in the future. How do you check for it?</p>

<p>And Leif, I consider anyone a king who put the time, effort and money in attaining a degree in any of the sciences. I respect science. Just wish I had the patience and concentration required to master one of them especially in the field of physics. I'm just glad I have the opportunity to talk shop with you guys with my non-science background. It's one of the reasons I hang out here at Photo.net instead of Facebook.</p>

<p>Oh wait, I wouldn't hang out at Facebook anyway if it was the last site on the internet, but I think you get the point.</p>

<p>Thanks for the very well thought out explanations.</p>

<p> </p>