Micro Four-Thirds, film comparison

[Adrian Bacon]

I agree with certain reservations. Viewing distance is one...

Viewing distance does play a role.

...and I also think there's a maximum size per format before sharpening detracts from the balance of a photograph. For example if you enlarge an image in editing software and concentrate on a fine detail like grass, it will sharpen up nicely but the image as a whole can appear over-sharpened in smoother areas, with what people have described as a visual mosquito whine.

This is why you selectively sharpen, meaning, only sharpen areas of interest. Just blindly sharpening the whole image leads to things like visual artifacts in areas that should be smooth (like sky). At the risk of sounding like a jerk, anybody who just throws a sharpen filter over the whole image before printing a reasonably large print does not know what they are doing. Just like with shooting with a small DOF field to draw the viewers attention, or using leading lines to draw the viewers eye, intelligently sharpening only the areas of interest has a dramatic effect on the image.

When I’m preparing an image for print and have it resized to the native print resolution for the size print it’s going to be, I routinely have multiple sharpening layers with layer masks masking off just the areas I want sharpened. Each sharpening layer is a different amount of sharpening, so not only do I selectively sharpen parts of the image, I also control how much sharpening is to be applied to each area. This way you can sharpen up the grass (using your example), but sharpen it less than the subject of the picture, and not sharpen the sky and areas that are not important or out of focus at all. I generally make the subject of the photo the sharpest, then progressively dial back the sharpening on other elements in the image the further away from the subject they are.

Image sharpening is a tool in the visual image toolbox just like any other photographic tool that you would use to manipulate the viewer viewing the image.

 

[Adrian Bacon]

I've never understood how dedicated interpolation software works. Sharpening, contrast, saturation and the rest I get, and the role they play in enhancing the perception of "sharpness", among other things. Does it add pixels based on the proximity of similar pixels to effectively make a larger file and offer more information?

Edit: I took the opportunity to look it up. It seem bicubic interpolation, which interpolates the nearest 16 pixels and is considered the best, degrades the image between 120 and 150% enlargement, which doesn't seem like a lot.

I’ve written a fair amount of interpolation software. What you get in photoshop and Lightroom is actually fairly rudimentary compared to what you could do. At its basic core, in simplified terms, interpolation is coming up with a value at a certain point along a line by using the values along that line on either side of that point as a reference. The simplest algorithms (like linear or cosine interpolation) only use one point on each side of the interpolated value. More advanced algorithms (like cubic, lanczos, etc) use multiple points on either side of the interpolated value to come up with the interpolated value.

Applied to image resizing, determining the directions of the lines and what samples to use on each side of each interpolated value is actually the hard part, as that has a very big impact on how much fine detail you retain. When you see something like bicubic interpolation, all that means is it is generically applying cubic interpolation in two directions, along the horizontal axis and along the vertical axis of the image being resized. For many images, depending on how much you are resizing, this is rudimentary, but effective.

If you want to get advanced, you do some image analysis to determine the direction of the edges in the image, and interpolate along the long side of the edges using a more sophisticated algorithm like centripetal catmull-rom. How many surrounding pixels (or the neighborhood size) you use can affect how intelligent you can be in determining the best direction to interpolate along.

Handling a Bayer array and turning it into a standard rgb image is interpolation at its core. It’s also unique because not only are you interpolating to get your final image, but you’re dealing with different amounts resolution between the colors. You have twice as many green pixels as you do red and blue. It’s quite fascinating.

 

[blockend]

When I’m preparing an image for print and have it resized to the native print resolution for the size print it’s going to be, I routinely have multiple sharpening layers with layer masks masking off just the areas I want sharpened.

That sounds like so much trouble. I rarely sharpen anything, even for print it's the lowest possible percentage. Digital images look over sharpened and over saturated to me straight out the camera, and I generally go the other way. I darken the highlights a little if a sky is blown, and drain half the colour out and that's pretty much it.

What you get in photoshop and Lightroom is actually fairly rudimentary compared to what you could do.

I'm still on Photoshop Elements 6 and I've never looked under the bonnet of most of that.

You have twice as many green pixels as you do red and blue. It’s quite fascinating.

I'll take your word for that. I don't think my Fujis work that way. Maybe my reluctance to play with the toolkit is why my M4/3 pictures are only good for 16 x 12?

 

[MattKing]

I'll take your word for that. I don't think my Fujis work that way. Maybe my reluctance to play with the toolkit is why my M4/3 pictures are only good for 16 x 12?

Does Fuji make a micro 4/3 product?

 

[faberryman]

Does Fuji make a micro 4/3 product?

It is not a member of the alliance.

 

[Adrian Bacon]

I'll take your word for that. I don't think my Fujis work that way. Maybe my reluctance to play with the toolkit is why my M4/3 pictures are only good for 16 x 12?

Fuji’s primarily use their x-trans color filter array which is even more green resolution relative to red and blue than a Bayer array. That one is even more difficult to turn into a standard rgb image.

Well, if you shoot raw and don’t do any sharpening at all, you can get quite a lot more detail pulled out. If you shoot joeg and think it looks too sharp, you should be able to go into the picture profile on your camera and dial down the sharpening. I primarily shoot Canon and always go with the neutral profile, then go in and turn the sharpening down as far as will go, then turn the contrast down as far as it will go, and usually dial the saturation back a couple of clicks. I shoot both raw and joeg, and for a lot of images it’s SOOC joeg with the modified profile, but for some, if I’m going to make a print larger than 8x10, I’ll pull the raw file in and really tune it up specifically for the intended output,

 

[blockend]

Does Fuji make a micro 4/3 product?

Not as far as I know. I have two Fuji cameras and a Panasonic. The Panny is newest, the Fujis are 2012 models.

If you shoot joeg and think it looks too sharp, you should be able to go into the picture profile on your camera and dial down the sharpening.

I shoot jpeg and Raw, manually expose and tweak the profile to get the jpegs where I want them. If there's a good shot I'll process it a little, but prefer to make my photos look the same. Some of the newer simulations and profiles are very good, in camera and software.

I primarily shoot Canon and always go with the neutral profile

I had a Canon until last year and kept the lenses, the colours are very good. I used a very flat default profile in DPP software and it's probably the nearest thing to a colour negative film look. Black and white simulations are getting better all the time, though I prefer to think of them as digital finally catching up with how a monochrome photograph should really look.

if I’m going to make a print larger than 8x10, I’ll pull the raw file in and really tune it up specifically for the intended output,

That makes sense. I should really spend more time processing, though when I do I often go back to the beginning and think I should have let things be!

 

[Adrian Bacon]

That makes sense. I should really spend more time processing, though when I do I often go back to the beginning and think I should have let things be!

Post processing is a lot like salt. A little bit goes a long way, and if done just right, makes all the difference in the world. You can get by with no salt, but just a little done just right is so much better.

 

[trendland]

but I maintain it can produce a totally credible and visually balanced image at 16 x 12", perhaps a little bigger, which is all I ever expected of a 35mm print.

From same point of view a friendly photographer stated years ago (decades) :

"There where you have the need of highest quality you have to shot sheeds! And 4x5 is the smallest
one - but for the professional work you have to come with 5x7 (or 8x10)!

There where you don't have the need of highest quality you should set priority with 120 film format! It is much much cheaper from films and from equipment and from weight!

And 35mm was the all time favurite to amatheuric shooters! "

Time has changed : No digital substitude for high resolution with 8x10 film !

There is a todays hart battle between highliest priced 150MP digital cameras and max. 4x5 inch output!

120 film is replaced from digital systems with big sensors!

And 35mm film can be replaced from full sized sensor systems!

But I of course have problems to regard anything smaller sized (incl APS C) as "Non amatheuric"

If you feel fine with micro four thirds you may feel fine!

But all the time within the last decades amatheuric photography had the need of professional
equipment! From were is the change today? Why you give such lot of money for digital?

So if it is quite clear : "mirco for third" has not the potentional to survive? To me it was clear from
date of introduction by Olympus!

So every lens you will buy next is a full lost investment! There is a new battle today :

Smartphone vs all digital systems of smaller size ! And all manufacturers of digital camera equipment will be horrable massacred!

There is the next upcoming niche = Nikon Canon with 120megapixel!

Everything below (incl. todays full frame 24MP class) will be worthless electronic scrap within
3years from now! But you are unable to learn folks! Because you hear the sound of advertising
sirenes!

with regards

PS : Odysseus sailed past the siren islands - you should avoid small sized sensors like
micro four thirds !

 

[warden]

My conclusion was that if you require absolute detail in a context like landscape, where the ability to discern and describe individual pieces of foliage and stones at distance is a way in to the subject matter, 16 x 12" is a useful working limit. Which happens to coincide with my experience of 35mm film over the years.

I think your conclusion is reasonable, and if your habit is to print no larger than that anyway and you like the way the lens/sensor combination renders, then micro four-thirds makes a lot of sense for quality, portability and price.

 

[Ste_S]

It isn't much of the daylight. If it is November in Quebec. If you take pictures before and after work you have less than two hours daylight window, for the rest you are dealing with dark street. I'm not machine gun shooter, my film lasts two days if not more. But I do look for pictures all the time. Day and evening, indoors and outdoors.
Also, if you are into street, candid photography you want it at least at 1/250 and f5.6.

I hope it rings the bell now? .

You're kinda missing the point of m4/3 cameras - which is the large depth of field. For the same depth of field you're getting at f5.6 on 35mm, you're shooting at f2.8 on m4/3. So for your 35mm ISO 3200 you're shooting at ISO800 on m4/3.

Crop sensors make a whole lot of sense for street photography, heck with a phone camera everything is in focus.

 

[Ko.Fe.]

You're kinda missing the point of m4/3 cameras - which is the large depth of field. For the same depth of field you're getting at f5.6 on 35mm, you're shooting at f2.8 on m4/3. So for your 35mm ISO 3200 you're shooting at ISO800 on m4/3.

Crop sensors make a whole lot of sense for street photography, heck with a phone camera everything is in focus.

I read about this theory. This is why I'm still keeping original Oly mft pancake and OVF. No camera, yet.

I'm just not sure how twice smaller sensor is translatable to three times lower ISO.

 

[trendland]

Crop sensors make a whole lot of sense for street photography, heck with a phone camera everything is in focus.

That's indeed true concerning shots with the need of much deep of field! What I remember from
people with APS C DIGITAL is that whimpering with portrait shots and the need of a new lens
with f 1.2 !

I read about this theory. This is why I'm still keeping original Oly mft pancake and OVF. No camera, yet.

I'm just not sure how twice smaller sensor is translatable to three times lower ISO.

It is translated in regard of deep of field - you understand KO.Fe.? But you may be right in concern
of croping factor - is it x2 indeed with APS C vs 24x36? I guess yes!
But deep of field is concerning to the lens design and f- stop AND format - so Ste_s can be right!
(If he found out with bis cameras - lenses)

with regards

PS : NO APS C is confortable for me!

 

[Alan Johnson]

A test chart, photographed from the correct distance lit by off camera flash..
Top a 35 mm Canon EF 50 mm @ f8 , Delta 100, Pyrocat HD scanned with Plustek 8100 @7200 dpi
Bottom mft Lumix GH2 . 20mm f4 same overall subject width.

 

[faberryman]

At this point I am asking myself who cares if m43 has worse, the same, or better resolution than 35mm film. Why would that be a relevant criteria for anything?

 

[trendland]

A test chart, photographed from the correct distance lit by off camera flash..
Top a 35 mm Canon EF 50 mm @ f8 , Delta 100, Pyrocat HD scanned with Plustek 8100 @7200 dpi
Bottom mft Lumix GH2 . 20mm f4 same overall subject width.

View attachment 223145

View attachment 223146

Let me ask you Alan : Why digital sensors show these all time phantastic resolution/quality?
Would you agree with my answer : Because most comparisons between are made from scanned
workflow and such test ALLWAYS
give answer about resolution of scanning unit vs sensor resolution - so digital sensors can not
lose that battle!

with regards

PS : Would you agree?

 

[Theo Sulphate]

...
Because most comparisons between are made from scanned
workflow and such test ALLWAYS
give answer about resolution of scanning unit vs sensor resolution - so digital sensors can not
lose that battle!
..

Yes. On dpreview.com so many people believe film is no better than 6 MP because on their monitor they are constantly comparing (poorly) scanned film to a raw image from their digital camera's sensor.

 

[trendland]

A test chart, photographed from the correct distance lit by off camera flash..
Top a 35 mm Canon EF 50 mm @ f8 , Delta 100, Pyrocat HD scanned with Plustek 8100 @7200 dpi
Bottom mft Lumix GH2 . 20mm f4 same overall subject width.

View attachment 223145

BTW : This has to tell us much about smal sensor abilities to produce resolution!

with regards

PS : Because concerning a 300bucks scanner was able to win that !

 

[Alan Johnson]

The film results are scanner limited, under the microscope I can see !00 lppm.More careful worker Henning Serger reported even higher resolution for Delta 100.
If the film is printed with an enlarger and a good lens, better results than scanning can be obtained.

 

[trendland]

The film results are scanner limited, under the microscope I can see !00 lppm.More careful worker Henning Serger reported even higher resolution for Delta 100.
If the film is printed with an enlarger and a good lens, better results than scanning can be obtained.

Yes Alan - Henning is a todays hero indeed! Because he made a final end with that test hocuspocus
we saw over a full decade!

I well remember a Nikon D800 test with Delta100 up to 4x5 inch (won by Nikon)
caused from flatbed scanning!

with regards

 

[faberryman]

The film results are scanner limited, under the microscope I can see !00 lppm. More careful worker Henning Serger reported even higher resolution for Delta 100. If the film is printed with an enlarger and a good lens, better results than scanning can be obtained.

Saying film has more resolution requires a context. Since so many film users today scan and print digitally, the comparison is helpful, if only to encourage people to avoid scanning and instead optically print. If you don't do it yourself, good luck finding someone who will, particularly for color. I see little advantage to shooting film and scanning. If you are going to all the trouble to shoot film, follow up with an all analog process.

 

[trendland]

Yes. On dpreview.com so many people believe film is no better than 6 MP because on their monitor they are constantly comparing (poorly) scanned film to a raw image from their digital camera's sensor.

I guess Theo - if digital have not been pushed that extreme between 2002 (2MP - 4MP)
and 2005 (8MP class digital cameras) amatheuric shooters would have not made that extreme
decision for digital!
Remember 2005 : all camera manufacturers gave up the production of film based models!
with regards

 

[trendland]

Saying film has more resolution requires a context. Since so many film users today scan and print digitally, the comparison is helpful, if only to encourage people to avoid scanning and instead optically print. If you don't do it yourself, good luck finding someone who will, particularly for color.

It is possible Faberryman (but exclusive and expensive) - like everything concerning film!

with regards

 

[warden]

Saying film has more resolution requires a context. Since so many film users today scan and print digitally, the comparison is helpful, if only to encourage people to avoid scanning and instead optically print. If you don't do it yourself, good luck finding someone who will, particularly for color. I see little advantage to shooting film and scanning. If you are going to all the trouble to shoot film, follow up with an all analog process.

The advantage to shooting film and scanning is simply flexibility. With film you can optically print the important negatives if you want/need maximum quality, and scan the others for less demanding use.

 

[alanrockwood]

The paper "A Pixel is Not a Little Square" is a great read for anyone looking for more insight into how that sort of data can be processed, and why some of those things work.

I found the paper and started reading it. I have not made my way through the entire paper yet. The premise of the paper is very interesting and thought provoking. It is also wrong in at least one important way.

The premise of the paper is that a pixel is a point, not a square. The problem is that the concept that a pixel is a point does not map onto any physically realizable image acquisition system because there is no such thing as a point detector in the physical world. All detectors have size.

Furthermore, although one could approximate an array of point detectors by using very small detectors that are separated from each other by a distance that is large compared to the size of the detector, such a scheme would be inefficient with respect to collection of light because most of the photons would fall on dead zone between the detector elements. Such a scheme would, I believe, also be more subject to aliasing than a closely packed detector array.

One could therefore write a counter-paper with a title something like "A pixel is not a point...."

The paper is also a simplification or distortion in another way. It defines and image in terms of a rectangular array. This ignores the fact that other geometries are possible, and some of them have advantages in certain respects. For example an array of hexagons is known to be a more efficient representation of certain image types than a square array. A wikipedia article on hexagonal sampling explains that a hexagonal sampling scheme requires 13.4% fewer samples to fully recover a circularly bandwidth limited signal. This can be very relevant to photography because a conventional lens produces a circularly bandwidth limited signal. Also, generally speaking a hexagonal array is also better able to deal with images with curved boundaries.

One could ask whether a Bayer-type sensing scheme can be efficiently mapped onto a hexagonal array. I don't know the answer to this, but in any case a hexagonal array would be a natural fit to a Fovean-type sensor.

These comments are not mean to denigrate the significance of the paper, but only to point out a few of the limitations or simplifications inherent in the paper.