Resolution - Digital versus 35mm Film

[jaapv]

Advertisement (gone after registration)

I can see that it might be a problem if you are a corporate photographer and have to photograph pin-striped suits all the time, or in similar applications. But I don't think that to require all images to be soft to avoid some situations is a very elegant solution either. One step back may do the trick, a slight defocus if moire occurs, use an old soft lens whatever. All similar solutions as a matte screen in front of the sensor. Moiré is rare enough in general photography.

Although - bird's feathers..(not too bad usually..) But a pre-war Junckers aeroplane? forget it...

[Advertisement]

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

[dhsimmonds]

Many of the MF back manufacturer's offer an optional and removeable AA filter. Why modify all images when probably only very few will need correction?

 

In the three years and many, many thousands of images that I used a DMR back without AA filter, I never used the optional moire switch.

[tgray]

Part of the problem with aliasing is you don't actually always know what is real detail what is not. It doesn't always end up looking like crazy swirly patterns in a shirt or maze like patterns. Zeiss just put up a great article about MTFs and imaging with some great examples here. Read if you feel like it.

 

Anyway, near the end, there are some examples of an illustration of a microscope, and 24mp and 12mp images of the illustration. Even though they don't really talk about the need for AA filtering too much, the 12mp images of the microscope show some great examples of moire, with detailed patterns that look completely real, yet are nonexistent in the original. Parts of the illustration look cross-hatched when in fact its simple hatching. And if someone presented this image, most people would assume that the camera was picking up real detail and not aliasing. There's also a nice example (last photo) how diffraction effects change resolution. Not stated, but visible, is how the aliasing goes away in the diffraction limited images. Not as sharp, but truer to the original scene.

 

They make some very nice comparisons between the 24 and 12 mp images. There are very real gains from bumping up resolution. Mind you in all of this, they never once talk about dropping the AA filter. Oh actually they do. Once. It's in reference to cell phone cameras:

"And yet the photo-module can do without a low-pass filter which is, in a way, incorporated into the lens since the sensor performance is so close to the physical limits of the optical system."

 

AA filtering is good. You give up some sharpness yes. You also prevent aliasing. The solution to this compromise is to bump up the resolution, not get rid of the aliasing filter. If we get it up high enough, *without* compromising other parts of the data acquisition, then we can do without a physical filter and rely on other features of our optical systems to introduce the necessary low pass behavior.

[pgk]

The whole expose to the right thing is founded on silly thought processes.

 

Well I 'expose to the right' and it works! Photographic theory and practice need to be utilised together and ALL factors figured in. I am perfectly satisfied that its easier to 'expose to the right' even to the very edge of clipping in order to produce better mid and shadow tonality, having tried it. But this is for the type of image I shoot.

 

As to AA filters, well after using my M8 for 18 months I can think of only one shot where I could detect moire, and the images it produces are very 'crisp' indeed and require little or no usm if utilised at full resolution. My Canon's always require usm and I personally do prefer a quicker workflow where that I don't have to use a usm twice - once to deal with the AA filter and once to finalise the output file - but this is inevitable with AA filter.

 

These are just my thoughts which work for me. Other people obviously have other viws and its a case of whatever works best for each of us.

[tgray]

Well I 'expose to the right' and it works! Photographic theory and practice need to be utilised together and ALL factors figured in. I am perfectly satisfied that its easier to 'expose to the right' even to the very edge of clipping in order to produce better mid and shadow tonality, having tried it. But this is for the type of image I shoot.

 

I agree that it has positive benefits. But the reasoning behind doing it is wrong. It's like saying eating lots of vegetables will make you healthier because they keep the evil spirits at bay. They are healthy for you, but it has nothing to do with evil spirits. And you're point drives it home - it gives you better shadow tonality, not better highlight tonality, which the faulty reasoning tells you it should give you.

[stunsworth]

It's like saying eating lots of vegetables will make you healthier because they keep the evil spirits at bay. They are healthy for you, but it has nothing to do with evil spirits

 

But that's better than saying "Vegetables make me better, but I don't know why, so I won't eat them". The important thing is is that they make you better, the explanation is secondary. Do I need to disassemble my car engine before I can drive it? No. Actually I think it's the re-assembly that would prove difficult <grin>.

 

Do I need to sharpen M8 images? No. Do I need to sharpen 5D images? Yes. Why? I don't really care, I prefer the M8 images to the ones from the 5D.

[tgray]

Advertisement (gone after registration)

But that's better than saying "Vegetables make me better, but I don't know why, so I won't eat them". The important thing is is that they make you better, the explanation is secondary. Do I need to disassemble my car engine before I can drive it? No. Actually I think it's the re-assembly that would prove difficult <grin>.

 

Do I need to sharpen M8 images? No. Do I need to sharpen 5D images? Yes. Why? I don't really care, I prefer the M8 images to the ones from the 5D.

 

Fair enough The real explanations aren't all that hard though, so why not use them and avoid creating a religion out of something? I personally DON'T shoot with the highest resolution sensor out there, or with one at all. I like Tri-X and TMAX 3200, because of how it looks. I mean, heck, Tmax 3200 has all the resolution of a crummy 2 mp sensor probably. At the same time, I'm not going to sit here and argue with you how Tri-X out resolves a 12 mp sensor because the grains of the film aren't on a fixed grid like pixels in the sensor.

 

And I think you'd like the images from a 40 mp sensor with proper anti aliasing going on compared to your M8, even if you just then down-rezzed the resulting file to 10 mp. It'd be just as sharp (probably sharper) and you'd have no moire. But maybe you wouldn't. What do I know?

[AlanG]

I can see that it might be a problem if you are a corporate photographer and have to photograph pin-striped suits all the time, or in similar applications.

 

Repeating patterns is very common in architectural and interior photos - grids, grates, screens, etc. on the outside, all kinds of patterns in cloth and other materials on the inside. You can't just step back either - as you'll hit the wall. Besides these patterns are not in a perpendicular plane as with a person's suit and are in planes at various angles.

 

Even with an AA filter my original 11 Mpix full frame camera occasionally produced moire. This got better in the 12.5 megapixel model. (I can't say if this was due to increased resolution, a different AA filter or better processing.) After a few thousand images with the 21 megapixel model, I haven't seen moire yet.

 

DXO applies its own lens sharpening automatically when doing raw conversion with an option to use unsharp masking instead or in addition. C1 by default has a pretty high USM setting, but even if you turn it off it looks like it does some kind of sharpening or USM. So sharpening doesn't add to the processing time. From what I can recall, the 21 megapixel images have quite a bit of detail even without sharpening.

 

Even with lenses that are not considered top notch (Sigma 12-24,) there is an improvement in resolution between images from the 12.5 megapixel camera and images from the 21 megapixel camera. So you might want to consider what an M lens could do if the M8 camera had a 20-30 megapixel sensor.

 

I am thinking 40 megapixels in a full frame camera is about the usable limit for some time. That would give about the same sensor density of the current APS 15 megapixel cameras. So if the lenses show a difference from 10-15 at that format, they also will improve going from 24-40 megapixels on full frame. Beyond that will take 80 megapixels for 50% more resolution. I'd think that would produce diminishing returns considering the lenses and the hassle of having such big files. Plus how many shots need this much detail? And small steps beyond 40 megapixels will be fairly inconsequential.