Has the day of super high pixel counts passed

[jaapv]

Advertisement (gone after registration)

With this in mind, the video heavy features of the Leica SL suddenly make Leica management appear to be very, very forward-thinking.

 

 

Not THAT forward thinking.

 

Dr. Paul Wolff, the well-known photographer from the 1930ies and closely linked to Leitz  made some of  his famous skiing photographs using a movie camera, despite implying (not saying ) they were all taken with a Leica - early marketing-.

Nevertheless well worth looking at his work, btw.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[pgk]

But still, a trained eye will see the differences in colour and tonality even in smaller images, it's up to you and your clients if you justify or see it.

 

Unfortunately I will absolutely guarantee that this is not so. My undergraduate project involved an analysis of two systems (one of which was Leica) from utilisation of cascaded MTF data and a visual comparison of 10" x 8" prints (deliberately small). Despite the MTF data indicating the Leica based system to be significantly superior, viewing the 10" x 8" prints showed no discernible difference (and the viewers were looking for differences and would have been well able to see them if they existed as all were involved in imaging and photography). Downsizing a file cannot show differences where the data does not exist I'm afraid. Discarding data to downsize it does just that - if this were otherwise then huge files downsized to internet viewable jpegs would illustrate their superiority - but they don't.

[Paul J]

Unfortunately I will absolutely guarantee that this is not so. My undergraduate project involved an analysis of two systems (one of which was Leica) from utilisation of cascaded MTF data and a visual comparison of 10" x 8" prints (deliberately small). Despite the MTF data indicating the Leica based system to be significantly superior, viewing the 10" x 8" prints showed no discernible difference (and the viewers were looking for differences and would have been well able to see them if they existed as all were involved in imaging and photography). Downsizing a file cannot show differences where the data does not exist I'm afraid. Discarding data to downsize it does just that - if this were otherwise then huge files downsized to internet viewable jpegs would illustrate their superiority - but they don't.

For all your supposed credentials you speak a lot of nonsense. Huge files downsized to internet viewable jpgs is not the same as supersampling. If you can not see the difference in "such simplistic solutions" then I can't be bothered, like you say, reading further or entertaining in conversation. Can you please at least thoroughly test these cameras and workflows before guessing about their output? and telling people who know otherwise that they are wrong? To suggest more MP does not increase resolution is beyond laughable.

[Guest JonathanP]

I'd really like to get to the bottom of this dicotomy:

 

+ S images have a "walk in" feel to them -- way more dimension than anything off my M digital.

Maybe it's the superior S lenses, maybe it's the S' larger sensor .... maybe it's a combo of those things given the camera was built for digital from the ground up .... but the S offers a look of 3D'ness that eclipses the M, with files that stretch to print sizes and which are more malleable (and don't fall apart) at prints that are far larger than the 24mp --> 38mp headline would infer.

 

 

Downsizing a file cannot show differences where the data does not exist I'm afraid. Discarding data to downsize it does just that - if this were otherwise then huge files downsized to internet viewable jpegs would illustrate their superiority - but they don't.

 

I've seen a lot of S and M images on the web and can clearly see the difference as Jon describes, but I understand the argument that Paul is making. So is it just optics differences between the M and S lines, or does the oversampling of the bayer array show a benefit of the higher MP sensor?

 

Jonathan

 

ps there was an interesting article by Tim Parkin a while ago: https://www.onlandscape.co.uk/2010/11/the-mysterious-case-of-the-missing-berries-and-other-stories/ where he demonstrated the very noticeable loss of red berries in a photo taken on a 5DmkII compared to 5x4 reversal film. I think this is probably a result of the bayer array reducing red resolution and showing that 21MP isn't nearly enough to approach the lens resolving power for colour images. You need to oversample to compensate.

[pgk]

It is extremely frustrating when reality fails to coincide with theory. Unfortunately, as I have read elsewhere and do actually believe, it is rarely reality that is wrong. Out of curiosity, at what size do these differences actually become apparent?

[Guest JonathanP]

I can see the 'S' look at normal web sizes (~1000px) - as one example, there's a large thread of mixed M and S images (nearly made the mistake of typing that the wrong way round...) at Fred Miranda and if you scroll up from the bottom of a page you can often see the images before finding out which camera is used. The S images are very easy to identify. Now, one other possible contribution might be that the S photographers have better post processing skills or might be taking more care when making the images - who knows. I'm just curious really! 

 

Jonathan

[jip]

Advertisement (gone after registration)

Nikon does both races, the 'we don't need 40Mp cameras' and the 'we do need MORE than 40mp cameras' it's easy for them, they reach different mind sets this way

[saxo]

I....

ps there was an interesting article by Tim Parkin a while ago: https://www.onlandscape.co.uk/2010/11/the-mysterious-case-of-the-missing-berries-and-other-stories/ where he demonstrated the very noticeable loss of red berries in a photo taken on a 5DmkII compared to 5x4 reversal film. I think this is probably a result of the bayer array reducing red resolution and showing that 21MP isn't nearly enough to approach the lens resolving power for colour images. You need to oversample to compensate.

 

there are examples for black and white as well: Andre Giogoli published in his book about b/w photography a nice example. He showed a picture with vine stocks, the digital picture showed pure sticks, the analog one showed dots on the sticks, the vines. The digital image processing cleared the vines away....

[pgk]

I'd really like to get to the bottom of this dicotomy:

IMO the problem we have is 'simple' in that there is no 'simple' explanation. In my experience there are usually lots of factors involved when things don't work out as anticipated and the degree of complexity can be disconcerting. In the past we had far less control over images because we had to work in an analogue fashion and consequently we were very limited in what we could do with them, and much of what we could do require considerable understanding and experience of the processes involved. Today we (or the camera/software manufacturers) have far greater control and it is increasing. But to be able to utilise our knowledge of the interaction of lenses and digital sensors we also have to supply vast amounts of data from the lens to the camera so that we can 'correct' for inadequacies or undesirable attributes of both lenses and sensors and take into account the various parameters which affect them.

 

To explain how I see things; as an obvious example the digital Leica M utilises 6-bit coding to tell the camera which lens is being used. Fine, but really more data would be better - the aperture in use and focussed distance so that corrections can be made for both - ideally some appreciation of the image parameters might help - as varying subject distance will no doubt have subtle effects too. So to produce its high 'quality' images the Leica M rangefinder relies essentially on having superlative lenses whose characteristics are pretty stable - and a generalised correction works remarkably well (though some will no doubt disagree on occasion).

 

But push the MPixel boundaries out to huge sizes and any residual problems which cannot be taken into account by simple 6-bit coding will increase and become more obvious, and physical constraints will kick in too. So as far as I see it a simple increase in MPixels cannot on its own produce significantly better images once limits have been reached (and this will have an effect sooner in reality that at theoretical limits). So cameras which communicate with their lenses, and lenses which are designed with this in mind, and their characteristics understood so that they can be taken into account by the camera designers, will no doubt make use of all sorts of data and adjust the image accordingly. This is happening and there has been significant discussion about it here on the forum and some have not been too happy about it  .

 

So whilst we probably can already see the benefits of an integrated approach to camera and lens design already for some cameras this will probably become more obvious as time progresses. But simply increasing MPixels without taking into account a whole array of information won't improve things, it will simply reveal weaknesses. And it has done so on many older lenses already. Older lenses can be quite effectively used with digital cameras which can reveal many flaws and these can be used, sometimes to good effect. The higher the MPixel the more obvious such flaws become and its fun to try to optimise such lenses 'manually' in software although the balancing of each flaw soon illustrates why its difficult without appreciating the interaction between such flaws.

 

I could go on but .....

 

The original post asked 'has the day of the super high pixel counts passed?' and I suppose that the answer is probably not but equally I'd say that the assumption that simply increasing pixel count can 'improve' images is a flawed one. It will be interesting to see how manufacturers move forwards. Decreasing high end sales, the desire for more MPixels by those who simply like bigger numbers and the increasing complexity required to produce both cameras and lenses which will show small improvements whilst boosting the requirement for tighter manufacturing tolerances and quality control (does this strike a chord with anyone) is going to make for an interesting time. I may be wrong but perhaps Leica is well placed to operate in such a market given its expertise and obvious decisions to try what I would say are pretty  innovative approaches (multi format products with high interchangeability of lenses)?

[CheshireCat]

Nikon does both races, the 'we don't need 40Mp cameras' and the 'we do need MORE than 40mp cameras' it's easy for them, they reach different mind sets this way

 

 

Yeah, but it was "20 MP" a few years ago 

[jip]

Yeah, it just gradually increases with the buyers 'compensating factor'. The general opinion is always more/bigger=better. That will always stay.

[jip]

Anyway, I enjoy my S and M, and even analog scans of the R system. And just use it however I like. I never complain or want more or less resolution.

[thighslapper]

Anyway, I enjoy my S and M,

 ..... err .... there are other more discrete forums where you can discuss your private life ....... pm me if you want details ...... 

[jip]

..... err .... there are other more discrete forums where you can discuss your private life ....... pm me if you want details ......

If only Leica sold a D and the P system... Hahaha

[saxo]

Unfortunately I will absolutely guarantee that this is not so. My undergraduate project involved an analysis of two systems (one of which was Leica) from utilisation of cascaded MTF data and a visual comparison of 10" x 8" prints (deliberately small). Despite the MTF data indicating the Leica based system to be significantly superior, viewing the 10" x 8" prints showed no discernible difference (and the viewers were looking for differences and would have been well able to see them if they existed as all were involved in imaging and photography). Downsizing a file cannot show differences where the data does not exist I'm afraid. Discarding data to downsize it does just that - if this were otherwise then huge files downsized to internet viewable jpegs would illustrate their superiority - but they don't.

As long as you compare cameras with the same sensor size I agree with that, but different sensor sizes give different depth of focus when you use analog focal length,,, this influences the look of the picture.

[saxo]

 

 

To explain how I see things; as an obvious example the digital Leica M utilises 6-bit coding to tell the camera which lens is being used. Fine, but really more data would be better - the aperture in use ..

 

Just a hint: the leica M knows the aperture in use.

[jaapv]

Well, knows... More like an educated guess...

[saxo]

Another problem of the megapixel race: the useful minimum aperture will decrease and depth of focus decreases as well.

 

Example: diffraction starts at FF-sensor and 50Mpx at an aperture below 7. This means: hyperfocal pictures with an aperture of 8, or lower, will not benefit from high pixel counts, and, in addition, with increasing Mpx depth of focus decreases at given aperture. On the other side: for those who aim for the smallest depth of focus, high Mpx count is great.

[Advertisement]

Hallo,

Look here: Leica L-Mount

[saxo]

Well, knows... More like an educated guess...

Yes, but the guess is not bad, at least good enough for digital picture correction.

[Paul J]

Diffraction in 50MP 135 format is a non issue. Sure, it's there but the gains outweigh the losses and some of its effects are somewhat correctable in post. If someone needs maximum depth control PLUS diffraction free high resolution they need to move to a Tech Cam where they can shoot at more optimum apertures. Having worked with the 5DS R a bit lately, exposed correctly it is a superb camera and for it's intended use it's right on the money. It's a significant upgrade in resolution from the 5D MKIII and has better colour and tonality, no question. It trails the sony sensors for Dynamic Range but when exposed correctly and well lit it's very, very good.

 

I have no idea where this assumption stems that people are saying 'more pixels solves every thing'. I have read no one suggest such a thing or ask for more pixels at the expense of other qualities. While I'm not convinced Leica is yet capable of making a high res camera (by today's standards), it is the future, like it or not.