ThePhilips
Leading Member
That should serve as a reminder that we should get best out of what is available now. If we can afford it, of course.
Let's keep the "pity" under for the time being - while we enjoy the cameras and the glass.
-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
E-5: Moire and Megapixels?
- Thread starter Dorus
- Start date
Let's keep the "pity" under for the time being - while we enjoy the cameras and the glass.
I have done a few billboards (with E3 actually) and the file-supplying specifications were about the same as an A4. 10MP is fine.
The most demanding are smaller posters that hang on a wall, or for example pictures printed on a large size for a trade-fair or so... close viewing-distance. I have done a few of those as well (more than the billboards) and in some instances I had to stitch a few shots to get the quality right.
The most interesting one I have never seen in real life, a reportage picture of a large and very special construction-site, it has been printed wall-sized for on a trade-fair. Client was very enthusiastic, I wish I had seen it myself... I suspect that if the file is perfect, the result will still be pretty good at that size.
;-)
Lourens
Leica M8/M9 lack AA filter altogether, sharpness is outstanding, moire not a problem.
Poor engineering solution?
How do you know what AA filter is in E5 and why is it mismatched?
TIA.
Poor engineering solution?
How do you know what AA filter is in E5 and why is it mismatched?
TIA.
bobn2
Forum Pro
very good lenses
moire is a substantial problem. You see lots of M8/M9 shots with really visible aliasing.
Yes, a higher MP sensor with the appropriate AA filter is a far better engineering solution. All the detail, none of the aliasing. Leica, like Olympus, are the victim of circumstance. Originally they did without an AA filter for the DMR since there was no way to engineer one into that (no space behind the film SLR shutter. Similarly, building an AA filter into the M8 and M9 and keeping the M camera profile was impossible. So, naturally, their marketing department has made tha best of a bad job and emphasised the 'advantages' of no AA filter, and deemphasised the disadvantages. In Olympus case, the circumstance is not having access to a higher MP sensor. So they try to match the resolution advantage of manufacturers that do by having a 'weak' AA filter, and their marketing makes the best of a bad job. If the Leica 'no AA filter' is so good why did not Oly go the whole hog and lose that AA filter altogether.
All the publicity has been about it being 'weak' and 'weak' is a euphemism for mismatched. The results coming out show this, with clear aliasing in the images.
--
Bob
Rriley
Forum Pro
- Messages
- 21,846
- Solutions
- 1
- Reaction score
- 474
pretty much all cameras are pushed to higher resolutions these days and exhibit moire at extinction resolution. While there hasnt been a formal Resolution test for E5 yet I have no doubt this will be the same, just as it is incidentally for 50D and 7D
its my take that detail appears to be higher than its direct competition
it may well be Panasonics moire relief software which they have been working on for some years, probably inspired by Leica's efforts with the same. If you look at EPL1 files you see something of the same things happening
--
Riley
any similarity to persons living or dead is coincidental and unintended
bobn2
Forum Pro
Aliasing is just like noise, impossible to reliably distinguish from real detail and therefore impossible to remove once it's there. 'Noise reduction' and 'moire relief' try to do the same trick, distinguish between real detail and false detail and flatten the false detail. The problem is always that somtimes false detail looks like real detail to the prgram, and doesn't get flattened, but the eye still sees it, sometimes the real detail looks like the false detail and gets flattened. In the case of aliasing, a much better engineering solution is to have a high sampling rate, much higher than the maximum level of detail you're interested in, and then the frequencies above the level of interest can be cleanly removed, either simply by not viewing large enough that you can see them or by a softwar LP filter. That solution is both easier to engineer and produces better results. The sole disadvantage is that it uses more processing power and memory, but it's by no means clear that 'moire relief' sorftware is going to use any less processing power or memory than a sensor with high enough resolution to give equivalent detail when properly anti-aliased.
--
Bob
goblin
Senior Member
Sorry for the delay, here's the translation from the french forum (the Chasseur d'Images one).
Hello all
I am not supposed to be a great proponent of a brand which I don't own and from which I don't have neither camera bodies nor lenses.
Just to let all the possibly unhappy people here (E5 release) know that I played around with the full size files from the E5. An that the y are simply remarkable.
On the Russian site on the shots of dynamic patterns one can see that the 200 iso benefits from a complex treatment in raw, with (probably) touched DR. One can see in the 1600 iso files a clean resolution, despite the attenuated grain. In native isos, one can see an exceptional stability of nuances.
All this goes in the directions of Olympus' own talk:
You might or might not know that there's a company out there called Apical, which specializes between other things into software treatment at the raw level, and which has contracts with several digital camera builders.
I have shared in another thread that Canon does the same thing.
So you will understand that this "simple" 12Mp sensor could pull a lot of good surprises out of its hat, after a little trip in a good photographer's hands.
Of course, I am still extrapolating, as I haven't shared any info with any serious reviewer yet.
Some of you know that I am a professional in the printing business, and from this experience I can tell a good file when I see one.
The fake megapixel debate put aside - 12mp is enough in 99% of the cases. Except in sizes such as A1/A0 (and even there, it's still usable). And the 2x crop factor remains a plus for birders and such.
In short - I believe you can expect excellent surprises.
More to follow when I have the time.
Hello all
I am not supposed to be a great proponent of a brand which I don't own and from which I don't have neither camera bodies nor lenses.
Just to let all the possibly unhappy people here (E5 release) know that I played around with the full size files from the E5. An that the y are simply remarkable.
On the Russian site on the shots of dynamic patterns one can see that the 200 iso benefits from a complex treatment in raw, with (probably) touched DR. One can see in the 1600 iso files a clean resolution, despite the attenuated grain. In native isos, one can see an exceptional stability of nuances.
All this goes in the directions of Olympus' own talk:
- Pixels are treated individually
- AA filter almost inexistant, with a processing at the software level.
You might or might not know that there's a company out there called Apical, which specializes between other things into software treatment at the raw level, and which has contracts with several digital camera builders.
I have shared in another thread that Canon does the same thing.
So you will understand that this "simple" 12Mp sensor could pull a lot of good surprises out of its hat, after a little trip in a good photographer's hands.
Of course, I am still extrapolating, as I haven't shared any info with any serious reviewer yet.
Some of you know that I am a professional in the printing business, and from this experience I can tell a good file when I see one.
The fake megapixel debate put aside - 12mp is enough in 99% of the cases. Except in sizes such as A1/A0 (and even there, it's still usable). And the 2x crop factor remains a plus for birders and such.
In short - I believe you can expect excellent surprises.
More to follow when I have the time.
goblin
Senior Member
Follow up:
I would suspect that their collaboration (with Apical) has been deper this time, as one can see that the DR is already higher from iso 200. That makes me think that the algorithms are used from the beginning, not after the fact.
To answer the question - a "treatment at the individual pixel level" is a shady way of saying that intelligent algorithms are used at the raw level, directly on the signal from the amplification circuits. It sounds better than saying "the image is retouched". Psychologically, we are still afraid of the ghosts of the past, from the times of the film vs digital battle. This battle wasn't that long ago, but the debate is crooked from the beginning. Because film was the same "liar" as the "electric" photography. In film, the slightest chemical step aside from the path resulted in a chromatographic massacre.
But still - the scars of the film vs digital dispute are still deep, and manufacturers still hate saying that they retouch the images at the basic level. It causes almost metaphysical problems for the users.
Which goes completely in the wrong direction, as all this efforts to RETOUCH the image are made in the direction of GETTING CLOSER to the real thing. The raw signal itself is often nothing to write home about, its the opposite more often than not.
I would suspect that their collaboration (with Apical) has been deper this time, as one can see that the DR is already higher from iso 200. That makes me think that the algorithms are used from the beginning, not after the fact.
To answer the question - a "treatment at the individual pixel level" is a shady way of saying that intelligent algorithms are used at the raw level, directly on the signal from the amplification circuits. It sounds better than saying "the image is retouched". Psychologically, we are still afraid of the ghosts of the past, from the times of the film vs digital battle. This battle wasn't that long ago, but the debate is crooked from the beginning. Because film was the same "liar" as the "electric" photography. In film, the slightest chemical step aside from the path resulted in a chromatographic massacre.
But still - the scars of the film vs digital dispute are still deep, and manufacturers still hate saying that they retouch the images at the basic level. It causes almost metaphysical problems for the users.
Which goes completely in the wrong direction, as all this efforts to RETOUCH the image are made in the direction of GETTING CLOSER to the real thing. The raw signal itself is often nothing to write home about, its the opposite more often than not.
bobn2
Forum Pro
I don't think the issue of 'retouched' raw files is to do with 'scars of the film vs digital dispute'. The issue is very simple. The raw file contains all of the information which that camera's sensor and capture chain can give. Any subsequent processing removes information - it might remove noise (or aliasing effects) as well, but it will remove real information, and once it's gone it's gone. Any algorithm that can be run in-camera can also be run outside the camera, and doubtless there will be improved techniques in the future, or available from third parties which might remove the noise or aliasing at the cost of less real detail. However, if the camera has already got rid of that, your goose is already cooked.
--
Bob
KJSDESIGN
Member
if you would give a url, we could do it ourselves... google translate is fantastic
A good read with some familiar faces, where are you from originally?
--
http://www.flickr.com/photos/guillaumeserandour/
--
http://www.flickr.com/photos/guillaumeserandour/
goblin
Senior Member
Follow up:
About the "old sensor used" debate:
In an "old" sensor, there are often digital guts which are totally new, with totally new algorithms. New micro lenses, different AA filtes, etc, etc. And Olympus shared that they are new.
Which makes it a NEW sensor.
One can never predict the results with just one factor. What we, the users, see, is only the tip of the iceberg. IEnd products can be the fruit of extremely advanced research and work. Take the D3, and its follower - the D3s. Take the EXCELLENT sensor of the 50D, add pixels, and you have everyone concerned that the initial 15Mp were already a lot for it. Yet, the remodeled one gets rid of the red artifacts up to 12800iso, while even the 5mk2 still generates some even at 400 iso as long as you underexpose a bit !!! Where's the logic in that ? Well, it's in the treatments.
Oly and Nikon follow the same direction. To raise the quality bar in a given market segment, it's a better idea to work AROUND the sensor, with new technologies, than to bite stupidly in the same commercial line - more megapixels, always new sensors. Oly engineers did not bite the bait. If you want a world champion of low noise - go for FF. No need to fight in a direction which would hurt everything else.
So to stay at a high quality level at lower resolutions, it was a good idea to stay at 12mp, which is covering the best part of possible usages. In such a situation one can put its money where his mouth is - in R&D, in the direction of matured, intelligent products. It's a smart choice, function of what you can squeeze out of a sensor.
If they put high quality electronics around and on the sensor, it's already a very expensive adventure. High quality components are expensive - tension deduplicators ( sorry, not sure about the translation here ), amplification, lower working temperatures. Study of sophisticated AA filters. Then comes the software part. Software AA - Leica did it as well. Then intelligent raw treatment. Everyone knows Nikon does it, with good results. And there's still plenty of potential in this software, as Apical shows. Canon did use the same treatment, and despite the critics of people who anticipated bad results - results were excellent.
All this advancements are discretely and efficiently applied. It is evident that the frantic, mindless pixel race (more pixels without too much thought, as one can not invest on both at the same time, or at least - not everyone), the R&D guys now fish for the "little" pluses which add up. All the better for us users. It is clear that Oly did its homework right with that one.
And the reason I write this is because I can compare raw files, and am nicely surprised compared to E3, E620, D300 or 7D. Of course, I do stay with Canon, as I am part of the very few who genuinely need more megapixels for my line of work, but as far as quality is concerned - Olympus put the bar pretty high. I would actually gladly use Oly for animal images with the 2x crop and a 90-250mm, which would be equivalent to my Canons, but I am too used to a lens image stabilization system, so here I stay.
Besides, Olympus' goal is not, and should not be, to conquer limited and fixed markets such as the CaNikon Pro market. The important for them is to give a really good product to their own clients, and I am firm on that one - they did so.
Unless the jpg available on the web are modified or fakes, of course. But if they are real - they are the sign of a magnificent work. You guys are lucky to have this lens range and engineers who can deliver such a first rate amiral vessel.
About the "old sensor used" debate:
In an "old" sensor, there are often digital guts which are totally new, with totally new algorithms. New micro lenses, different AA filtes, etc, etc. And Olympus shared that they are new.
Which makes it a NEW sensor.
One can never predict the results with just one factor. What we, the users, see, is only the tip of the iceberg. IEnd products can be the fruit of extremely advanced research and work. Take the D3, and its follower - the D3s. Take the EXCELLENT sensor of the 50D, add pixels, and you have everyone concerned that the initial 15Mp were already a lot for it. Yet, the remodeled one gets rid of the red artifacts up to 12800iso, while even the 5mk2 still generates some even at 400 iso as long as you underexpose a bit !!! Where's the logic in that ? Well, it's in the treatments.
Oly and Nikon follow the same direction. To raise the quality bar in a given market segment, it's a better idea to work AROUND the sensor, with new technologies, than to bite stupidly in the same commercial line - more megapixels, always new sensors. Oly engineers did not bite the bait. If you want a world champion of low noise - go for FF. No need to fight in a direction which would hurt everything else.
So to stay at a high quality level at lower resolutions, it was a good idea to stay at 12mp, which is covering the best part of possible usages. In such a situation one can put its money where his mouth is - in R&D, in the direction of matured, intelligent products. It's a smart choice, function of what you can squeeze out of a sensor.
If they put high quality electronics around and on the sensor, it's already a very expensive adventure. High quality components are expensive - tension deduplicators ( sorry, not sure about the translation here ), amplification, lower working temperatures. Study of sophisticated AA filters. Then comes the software part. Software AA - Leica did it as well. Then intelligent raw treatment. Everyone knows Nikon does it, with good results. And there's still plenty of potential in this software, as Apical shows. Canon did use the same treatment, and despite the critics of people who anticipated bad results - results were excellent.
All this advancements are discretely and efficiently applied. It is evident that the frantic, mindless pixel race (more pixels without too much thought, as one can not invest on both at the same time, or at least - not everyone), the R&D guys now fish for the "little" pluses which add up. All the better for us users. It is clear that Oly did its homework right with that one.
And the reason I write this is because I can compare raw files, and am nicely surprised compared to E3, E620, D300 or 7D. Of course, I do stay with Canon, as I am part of the very few who genuinely need more megapixels for my line of work, but as far as quality is concerned - Olympus put the bar pretty high. I would actually gladly use Oly for animal images with the 2x crop and a 90-250mm, which would be equivalent to my Canons, but I am too used to a lens image stabilization system, so here I stay.
Besides, Olympus' goal is not, and should not be, to conquer limited and fixed markets such as the CaNikon Pro market. The important for them is to give a really good product to their own clients, and I am firm on that one - they did so.
Unless the jpg available on the web are modified or fakes, of course. But if they are real - they are the sign of a magnificent work. You guys are lucky to have this lens range and engineers who can deliver such a first rate amiral vessel.
goblin
Senior Member
I'm just translating
Keep in mind the guy is in France, where that debate might have been different.As for the signal treatment - I believe his point is that what is removed is clear junk anyway. Like a car with the same engine, but with a newer, low friction transmission. At least that's what I think he means.
Anyway. Here's the link:
http://www.chassimages.com/forum/index.php/topic,102493.0.html
Long story
In short - I am also French, between other things
Aliasing can be predicted with very high reliability. It just takes a bit of processing power and it can be removed very reliably.
bobn2
Forum Pro
I don't believe you're right. In any case, if you're going to use a lot of processing power is not a better solution a higher res sensor with a properly matched AA filter which won't have so much aliasing in the first place.
--
Bob
goblin
Senior Member
After screaming for years for a weak or inexistant AA filter, here we are suddenly asking for a "More pixels sensor + AA filter" combo. Olympus would go crazy if they tried half of the stuff preconised in this forum
Rriley
Forum Pro
- Messages
- 21,846
- Solutions
- 1
- Reaction score
- 474
coincidentally or not the processing power was introduced with E30, it needed that power for 'art filters' and the custom 'incamera' edit functions that camera carries. I had quipped at the time that the additional power might come in useful, and it seems it has...
--
Riley
any similarity to persons living or dead is coincidental and unintended
Have you even seen an old movie where the cart wheels seem to rotate backwards?
Ever played with a stroboscope to stop the motion of a rotating disk?
Its all about sampling. If your sampling rate is faster than what your are trying to sample you are not going to miss anything in between. If you sampling rate is much slower than what you are trying to sample then you are obviously missing what is happening between samples. Start sampling at about the same rate as what you are trying to sample and you are going to get funny things like moire, sound beats and wheels that appear to rotate backwards.
CD music is sampled at 44kHz, why? Because your ear can distinguish sound up to 20kHz in frequency. If you could hear all the way up to 100kHz, you would find CDs to be missing something, you would say that they sound SOFT.
Camera sensors are sampling devices, the higher the MP count, the higher the sampling rate. If you have a line pattern (black/white) in your picture that is much tighter than your pixel density then you are under-sampling...
Just like the cart wheel that appears to start moving backwards as your sampling rate approaches the RPM of the rotating wheel, moire is something that occurs when your sensor density gets close to the rate of the (black/white) line pattern in the picture.
How do you prevent this? By increasing the sampling rate (higher MP) or by filtering the picture to prevent (black/white) patterns that are faster(tighter) than the sensor.
This is what the AA filter does. It makes the picture soft above the sampling rate of the sensor and you must have it . If the AA filter is perfectly matched to the sensor, it will only remove what the sensor cannot capture to start with. You want and you need an AA filter.
Trouble is that AA filters are not perfect and they tend to filter patterns that the sensor can detect. This is a common "soft" complaint from pixel peepers! A weaker AA filter on the other hand will produce moire. But filtering is a necessary evil.
The trick: implement part of filtering in software.
This is what the E5 seems to have done.
BRAVO! But no matter how good the new AA filter is, the 12MP sensor is still under-sampling what OLY glass can resolve.
For those who are comfortable in advance math...simply refer to the Nyquist Theorem.
Ever played with a stroboscope to stop the motion of a rotating disk?
Its all about sampling. If your sampling rate is faster than what your are trying to sample you are not going to miss anything in between. If you sampling rate is much slower than what you are trying to sample then you are obviously missing what is happening between samples. Start sampling at about the same rate as what you are trying to sample and you are going to get funny things like moire, sound beats and wheels that appear to rotate backwards.
CD music is sampled at 44kHz, why? Because your ear can distinguish sound up to 20kHz in frequency. If you could hear all the way up to 100kHz, you would find CDs to be missing something, you would say that they sound SOFT.
Camera sensors are sampling devices, the higher the MP count, the higher the sampling rate. If you have a line pattern (black/white) in your picture that is much tighter than your pixel density then you are under-sampling...
Just like the cart wheel that appears to start moving backwards as your sampling rate approaches the RPM of the rotating wheel, moire is something that occurs when your sensor density gets close to the rate of the (black/white) line pattern in the picture.
How do you prevent this? By increasing the sampling rate (higher MP) or by filtering the picture to prevent (black/white) patterns that are faster(tighter) than the sensor.
This is what the AA filter does. It makes the picture soft above the sampling rate of the sensor and you must have it . If the AA filter is perfectly matched to the sensor, it will only remove what the sensor cannot capture to start with. You want and you need an AA filter.
Trouble is that AA filters are not perfect and they tend to filter patterns that the sensor can detect. This is a common "soft" complaint from pixel peepers! A weaker AA filter on the other hand will produce moire. But filtering is a necessary evil.
The trick: implement part of filtering in software.
This is what the E5 seems to have done.
BRAVO! But no matter how good the new AA filter is, the 12MP sensor is still under-sampling what OLY glass can resolve.
For those who are comfortable in advance math...simply refer to the Nyquist Theorem.
goblin
Senior Member
Thanks for the great explanation.
Which reminds me to ask - what is the resolution of a 12-60 for example ? Or of a SHG lens ?
What sensor would be needed to match / exceed them ? How many megapixels ?
Which reminds me to ask - what is the resolution of a 12-60 for example ? Or of a SHG lens ?
What sensor would be needed to match / exceed them ? How many megapixels ?
Similar threads
Keyboard shortcuts
- f
- Forum