If not that's kinda silly to say. As soon as the captured image appears in the camera's EVF I (usually) know. It's screamed at me with an ACR display of Fit Image which is usually like 16 to 20 percent. It reaches out and pokes ya right in the eyeballs with extreme force on the first PS scale displayed as Fit Image full-screen which for me is usually 50% display scale and 65% image scaling. After a few more scaling steps when the image size is below 1000 pixels it's sometimes difficult to tell. And of course it's fairly impossible to tell with shots being scaled down to around 800x600 and less (which is one reason I don't understand people who post that small). But absolutely no pixel peeping is needed to know if the lens is mediocre, sharp, or exceptionally sharp - once a person has shot with those grades of lenses. And that's just what FlyPenFly said too....Show more →
there are certainly lenses that look bad at fit to screen size, but most lenses i shoot with i wouldn't be able to tell any difference in sharpness at 1920 pixels long (lots of other differences pop out though). especially not in photoshop, whatever the display scaling it uses is makes everything look soft (lightroom is much better in this regard). in fact, as discussed earlier, some of my lower resolving lenses with high contrast look sharper downscaled by PS/lightroom than some of my lower contrast lenses but when you look at 100% you can see the lower contrast lens is obviously sharper.
Bifurcator wrote:
And of course again just like FlyPenFly said there too, you should optimize for both the original 100% and for the targeted output. Nothing else even makes much sense. You can't tell me you don't do that S. I know you do it - I can see it in your images and I think I read you saying as much in the recent past too.
i almost never do that unless i plan to print at or above 100%. i optimize my originals for downsizing, which is quite different than optimizing them for viewing at 100%.
sebboh wrote:
i almost never do that unless i plan to print at or above 100%. i optimize my originals for downsizing, which is quite different than optimizing them for viewing at 100%.
Optimizing sharpness at 100% will on many types of motifs inevitably create moire/jagged edges in standardized step.downsizing scripts which I assume many of us use. Usually on things like straight lines and similar, less so for nature/landscape. I have stopped sharpening at the 100% level (above default in ACR) if focusing was not missed. But in high ISO-/short DOF situations where noise reduction is needed, a slight counter action in sharpening is appropriate..
But there's a difference if using FF - crop - or "worse" MFT. Smaller sensor needs more sharpness care at 100%. Note, this is true given equal pixelcount, ie the FF will have much lager pixels.
I bet some will object, especially people using small sensors tripods taking still life where it is more than possible to - with top lenses - get amazing results on e.g. the 5N at the pixellevel. But the cropped Canon 7D I left a couple of years ago needed much more care at 100% to be happy with a downsized final result.
I guess I have super-eyes, a super-monitor, and a special super-EVF then because I can usually see how sharp a lens is on those devices (with my settings) without having to zoom to 100% or past (which is what I think the definition of pixel-peeping is - especially past 100%). About the only time I pixel-peep is when I'm evaluating a lens for the first time or if I think I was shaking too much and wanna know by how many pixels it's blurred.
If your sharpening enhancement step at 100% (typically just prior to a down-scale operation or the defaults applied on opening them) causes unwanted anomalies there's a special key on your keyboard which will Undo or step backward (or you can turn down/off the defaults). :P Likewise in PS there is a tool called Fade which I personally use after almost every operation that changes pixels in any way.…
Keep in mind that every single RAW converter in the entire industry by default will sharpen your RAW images after conversion so it's not just me - it's an entire industry that is thinking this is the best, most commonly desired, or most appropriate way to begin processing a 100% image.
wfrank wrote:
But there's a difference if using FF - crop - or "worse" MFT. Smaller sensor needs more sharpness care at 100%. Note, this is true given equal pixelcount, ie the FF will have much lager pixels.
I bet some will object, especially people using small sensors tripods taking still life where it is more than possible to - with top lenses - get amazing results on e.g. the 5N at the pixellevel. But the cropped Canon 7D I left a couple of years ago needed much more care at 100% to be happy with a downsized final result.
This is interesting. I actually get lower per-pixel sharpness with the Zeiss 35/1.4 on D700 than with the Panasonic 25/1.4 on OM-D. The look is of course different, but even with the higher pixel count and tiny sensor, the OM-D wins regarding pure per-pixel sharpness.
That would make the most logical sense to me too. But I said as much 2 years ago with image examples too, and several people totally freaked out and hated upon me with a terrible passion! At that time it was the sacred D700 vs. the Minolta A2. So I think it's true even at those extremes. Maybe now (some of) those people are no longer here we can have this discussion intelligently without all the hate?
Makten wrote:
This is interesting. I actually get lower per-pixel sharpness with the Zeiss 35/1.4 on D700 than with the Panasonic 25/1.4 on OM-D. The look is of course different, but even with the higher pixel count and tiny sensor, the OM-D wins regarding pure per-pixel sharpness.
I have to say that 100% crops from MFT usually look a bit more "digital" to me. Pixel level detail looks more processed and bokeh often doesn't have any "texture" which makes it look as if it was made with gaussian blur in PS (whereas bokeh on FF looks a bit more organic and usually has more structure/grain). That's purely on the pixel peeping level though, but for me it's the giveaway telling me that I'm looking at images shot with MFT. Maybe it's also a lack of tonal gradations that causes this different look and I'd expect digital MF backs to be superior to FF in this departement too.
AhamB wrote:
I have to say that 100% crops from MFT usually look a bit more "digital" to me. Pixel level detail looks more processed and bokeh doesn't have any "texture" which makes it look as if it was made with gaussian blur in PS (whereas bokeh on FF looks a bit more organic an usually has more structure/grain). That's purely on the pixel peeping level though, but for me it's the giveaway telling me that I'm looking at an images shot with MFT. Maybe it's also a lack of tonal gradations that causes this different look and I'd expect digital MF backs to be superior to FF in this departement too....Show more →
+1
After mild sharpening, I prefer the output of my D700 (+ ZF glass) to the NEX / OMD, its very pleasant to see and more natural. Agree on the bokeh part also.
Has anyone here care to move the AA sensor of the D700? I'd love to see the results very much..
AhamB wrote:
I have to say that 100% crops from MFT usually look a bit more "digital" to me. Pixel level detail looks more processed and bokeh doesn't have any "texture" which makes it look as if it was made with gaussian blur in PS (whereas bokeh on FF looks a bit more organic an usually has more structure/grain). That's purely on the pixel peeping level though, but for me it's the giveaway telling me that I'm looking at an images shot with MFT. Maybe it's also a lack of tonal gradations that causes this different look and I'd expect digital MF backs to be superior to FF in this departement too....Show more →
Yes, MF backs look far superior IMO especially the bigger format MF backs.
IMO, tonal gradations are tied to dynamic range and that's traditionally been a very big weak spot of MFT. The Sony sensor in the OMD is leagues better but IMO still a bit behind the best from the NEX-7.
On the flip side, you have a lot of practical advantages on OMD that the nex-7 doesn't have as well as a host of native lenses.
Makten wrote:
This is interesting. I actually get lower per-pixel sharpness with the Zeiss 35/1.4 on D700 than with the Panasonic 25/1.4 on OM-D. The look is of course different, but even with the higher pixel count and tiny sensor, the OM-D wins regarding pure per-pixel sharpness.
Not that I doubt your finding (I rarely do :-) but it would be really interesting to see an example in case you have some specifics in mind. It totally counters my experience with 5D2 vs 7D. The NEX 5N is better but I usually attribute that to better lenses and sensor than the 7D. It would also give some flesh to the term per-pixel sharpness.
FlyPenFly wrote:
Hmm would one say Moire is a sign of high per pixel sharpness?
Yes. Moiré patterns or aliasing in general are caused by interference between high spatial frequencies in the projected image and the pixel grid of the sensor. A strong AA filter will attenuate this but also the demosaicing algorithm has a strong influence on aliasing/moiré and pixel sharpness.
I'm not a lens expert but I do know that moire will be caused by *any* high-frequency information -- it does not have to be *accurate* high-frequency information.
IOW, you might see moire with a lens that doesn't actually have good sharpness. It could be doing various kinds of detrimental things to the image, and as long as the output contains sections of high frequency, it could moire.
FlyPenFly, that moire looks to me to be much larger than the minimum resolution of the sensor, so I don't know what's going on there. Also, the building texture is grainy but doesn't strike me as something that would cause moire.
Sure it's not a post-processing artifact? It could also have something to do with the de-mosaic processing. But I don't think it's caused by sensor aliasing.
Taylor Sherman wrote:
I'm not a lens expert but I do know that moire will be caused by *any* high-frequency information -- it does not have to be *accurate* high-frequency information.
IOW, you might see moire with a lens that doesn't actually have good sharpness.
That doesn't make sense to me. If you're going to see moiré there needs to be high contrast at a spatial frequency in the neighbourhood of the sensor's Nyquist frequency. It's the moiré/aliasing that makes those image features inaccurate. If those frequencies are blurred to begin with you won't see moiré.
Of course, downsizing and then sharpening can introduce heaps of moiré, but that's a different issue.
What I see in FPF's sample looks to me like the typical color moiré to me but it may be less or even absent in other RAW developers (with different demosaicing algorithms).
AhamB wrote:
That doesn't make sense to me. If you're going to see moiré there needs to be high contrast at a spatial frequency in the neighbourhood of the sensor's Nyquist frequency. It's the moiré/aliasing that makes those image features inaccurate. If those frequencies are blurred to begin with you won't see moiré.
Yes. Lenses do more than just straight blurring. They can have all sorts of outputs for a given input. Even a fairly simple low-pass filter can exhibit "ringing", where a single sharp (very high-frequency) transition can be transmuted to a wave-like series of lesser-frequency (but still high-frequency) bands:
What I believe is happening in the above picture: the texture on the building is grainy, which means that it contains high frequencies but in a "noisy" pattern, eg most frequencies are represented equally (this is the definition of noise, more or less). Aliasing on such an input would normally produce a similarly "noisy" output - eg it should still be grainy.
However, the lens as a low-pass filter, exhibits ringing and most of the high-frequency noise energy that is the texture of the building is getting moved to a lower frequency, which must be > the sensor's Nyquist rate. Because this ringing is nice and wave-like, it results in blocky aliasing artifacts at much lower frequency.
What I don't know, is whether or not this indicates that the edge of the lens's resolution abilities is near the sensor Nyquist rate. In other words, is it possible for a lens which has a resolution limit of N exhibit ringing at frequencies substantially greater than N? I believe the answer is yes.
