Bayer - Foveon comparison

Started Nov 23, 2008 | Discussions
Lin Evans
Lin Evans Forum Pro • Posts: 17,609
Re: Too many sharpening artifacts

Hi Erik,

They both have too many sharpening artifacts including the Canon DPP conversion. Even the SPP large has bad black sharpening halo's around the right side of light upright supports. It's odd that there are not similar strength white sharpening halo's on the blue sky background. Usually when there are bad black halos there will be corresponding bad white ones.

The Photozoom spline interpolation definitely has bad sharpening artifacts, but I don't believe these are a result of the Photozoom interpolation but rather that the original was oversharpened.

The 50D jpeg has less obvious sharpening artifacts but the DPP conversion has bad white and white halos surrounding all structures.

Having said this, I doubt that any of these artifacts would have any serious adverse effect on most prints. On the other hand if this image was destined to be printed at A0 size I would not want to print any of them. All sharpening should be left until interpolation is complete and only then would it be practical to determine the degree of sharpening or deconvolution required for an optimal print.l

See halos on the 50D DPP conversion below:

Lin

docmaas
docmaas Veteran Member • Posts: 6,482
Re: Some creatures are even simpler

Good point Derek. Actually I think the superior inherent separability of pixels in the Sigma is due more to the absence of the AA filter than the 3 layer chip. Despite the interpolation of bayer I think you're right that since each position should be just as well separated in the bayer as in the foveon, and if there are more of them the finer granularity should lead to better demarcation between objects. However, the AA filter blurs the separation of pixels on the bayer filter and that is what really gives them the fuzziness rather than the interpolation per se. that's why the kodak looks more like a foveon image or leica image than do other bayer images with aa filters. Likewise the new Sony A900 is producing, according to many, inherently sharper images than its competitors and again the credit is going to a "lighter" Aa filter.

Now that FF sensors have reached far enough to satisfy modern lens resolution the next improvements will be on making those FF sensors better in other ways than just more mpix and I suspect that AA filters will become more sophisticated and perhaps even removable in the near future on DSLRs as they already are on MF sensors/cameras.

Mike

derek324 wrote:
Some creatures are simpler and even easier to confuse. I tried, and
tried, and tried again, with exactly opposite conclusion.

BTW, as an electronics engineer I would always argue that better
'registration of data' should be expected from sensors placed next to
each other, as opposed to 3 layers of sensors underneath each other,
no metter what the sensors are supposed to 'catch'. I just do not
believe in technology being able to offer total transparency between
the layers, zero reflections, and lack of interference... but this is
of course flogging a dead horse and was discussed many, many times.

Just shows that different perspectives do exists

richard stone wrote:

Simple creature that I am, and as I often say, easily confused, when
I look at these tests, time and time again, the Foveon images look
crisper and clearer to me. No blurring. Each photo-site is
reporting, spatially, exactly what it registered.

...

-- hide signature --

People do not seem to realize that their opinion of the world is also a confession of character.

  • Ralph Waldo Emerson

MOD Kendall Helmstetter Gelner Forum Pro • Posts: 20,516
No comment on detail vs. resolution?

To me your whole post did nothing to talk about what Lin was showing - a very clear illustration of detail vs. resolution. As he said, edges appear smoother in the Canon but look at the right side of that chain link fence - the whole thing is a mush in the Canon shot, while it's easy to see the links is almost the whole area of the SD-14 shot.

That is the difference between varying detail, and constant detail as the SD-14 offers.

You can say that we are pixel peeping but that chain link fence would seem to me something very easily taken in by the eye in a print as it unnaturally vanishes in the Canon shot since you can make out so much of the left side of the fence....

 Kendall Helmstetter Gelner's gear list:Kendall Helmstetter Gelner's gear list
Sigma 70-200mm F2.8 EX DG OS HSM Sigma 8-16mm F4.5-5.6 DC HSM Sigma 24-70mm F2.8 EX DG HSM Sigma 50-500mm F4.5-6.3 DG OS HSM Sigma 85mm F1.4 EX DG HSM +4 more
oluv Senior Member • Posts: 1,973
Re: Resolution versus detail......

what are you guys doing here? smoothing the 50D image on purpose? or are you comparing the SD14 raw to 50D jepg?

here is what i get with a decent raw converter, i don't see the same blurring you show in your comparison:

 oluv's gear list:oluv's gear list
Panasonic Lumix DMC-GM1 Panasonic Lumix G Vario 7-14mm F4 ASPH Panasonic Lumix G 20mm F1.7 ASPH Panasonic Leica Summilux DG 25mm F1.4 Olympus M.Zuiko Digital 45mm F1.8 +1 more
oluv Senior Member • Posts: 1,973
my take

many in this thread claim the SD14 to be able to resolve more details despite of having lower resolution. it indeed resolves an impressive amount of details for a 4.7 megapixel sensor and it compares very well to the 50D, but by no means it can resolve more details than the 50D. i have looked at various parts of the image, but i have not found a single part where the SD14 showed more details than the 50D file.

you can clearly see that the 50D is able to resolve the ropes on the very top much better, i can count the exact amount of wire-meshes on the top of the ladder in the 50D file, in the SD14 file it is still recognizable as wire mesh, but the are not resolved as single wires anymore. and also look at the bird sitting on the top of the handrail is just not resolved in the SD14 file anymore

i attached an image. the 50D file was converted with raw therapee and sharpened very slightly to compensate for bayer interpolation. the SD14 was converted with the latest ACR and not sharpened (i do not have SPP). i didn't want to try any upscaling or downscaling as the scaled image would always have a disadvantage over the unscaled one.

 oluv's gear list:oluv's gear list
Panasonic Lumix DMC-GM1 Panasonic Lumix G Vario 7-14mm F4 ASPH Panasonic Lumix G 20mm F1.7 ASPH Panasonic Leica Summilux DG 25mm F1.4 Olympus M.Zuiko Digital 45mm F1.8 +1 more
DMillier Forum Pro • Posts: 21,163
Re: No comment on detail vs. resolution?

As it happens, I cropped this part of the image out when I printed so don't have any personally generated data to comment on. Obviously, software and processing technique make a difference at the pixel peeping level - just see the Raw therapee conversion posted below.

Kendall Helmstetter Gelner wrote:

To me your whole post did nothing to talk about what Lin was showing

  • a very clear illustration of detail vs. resolution. As he said,

edges appear smoother in the Canon but look at the right side of that
chain link fence - the whole thing is a mush in the Canon shot, while
it's easy to see the links is almost the whole area of the SD-14 shot.

That is the difference between varying detail, and constant detail as
the SD-14 offers.

You can say that we are pixel peeping but that chain link fence would
seem to me something very easily taken in by the eye in a print as it
unnaturally vanishes in the Canon shot since you can make out so much
of the left side of the fence....

-- hide signature --
DMillier Forum Pro • Posts: 21,163
Re: Inferiority Complex

No one forces anyone to participate!

miketuthill wrote:
It's fun the first 656 times. By the 657th it starts to get a little
tired.

mike earussi wrote:

Comparisons are fun, that's why we do them.

-- hide signature --
DMillier Forum Pro • Posts: 21,163
Re: Bayer - Foveon comparison

Once we have played with all the settings, A3 prints from these cameras seem very similar to me. Colour differences, yes, but overall remarkably similar - which is hardly surprising as image quality from most cameras has reached a high quality.

I still don't see this extra reported realism from Foveon when I compare exactingly prepared identical pictures. It's there in 100% screen views of course, but in prints all looks remarkably similar.

I now have about 20 large print side by side exact comparisons between the Sd14 and high resolution bayer and frankly it is quite difficult for non afiliated lay people to pick which print came from a "special" camera.

There is no doubt in my mind that the Foveon does have remakable detail for its pixel count but when you shoot side by side prints I'm not so sure the claims of special qualities stand up.

It seems to me that by and large the specialness claim derives from people looking at superb Foveon images rather than side by side comparisons. It is hard to tell anything without a reference to compare to. Yes, if you see a diet of excellent Foveon images, you will be tempted to ascribe that to qualities of the sensor. Maybe, there are just a higher proportion of photgraphers using Foveon sensors who care about these things enough to take special care, so there is a higher proportion of excellent images from the cameras? Just a thought.

I do like the SD9 output though. Its colours are sometimes wacky but it brings a charm of its own to image making - a bit like Velvia, which is not a very good film in realism terms but in the right place makes superb looking images.

richard stone wrote:

Simple creature that I am, and as I often say, easily confused, when
I look at these tests, time and time again, the Foveon images look
crisper and clearer to me. No blurring. Each photo-site is
reporting, spatially, exactly what it registered. And it's not just
the lack of blurring, it's no interpolation at all. Color is a
different issue. Some color issues can be corrected. Blurry images
can be sharpened, it's true, but the result is often an unreal
sharpness.

The Foveon image seems more photographic, more of an accurate report
of what the lens provided as information and put on the sensor.

Is this going to show up in a print? I think the answer is yes, and
unless we are talking 4x6 inches, the Foveon will be better. But
most people don't care, and those that might care may not be looking
for or even notice the crispness and clarity that I believe is
present.

Early tests on the Outback site reported that Foveon sensors did
better on color images than Bayer, and the real world is in color.
Mike Chaney found the same thing. Yet when we get to these tests,
the tests generally come down to comparing non-colored lines against
a non-colored background, or very slightly colored tests. It's ok. I
just look at the images as posted and go, yep, the Foveon one is
crisper. I don't need to make it wall size to see that.

Thank you for posting your pictures and for doing this test.

Richard
--
My small gallery: http://www.pbase.com/richard44/inbox

-- hide signature --
DMillier Forum Pro • Posts: 21,163
Re: Some creatures are even simpler

I'd be careful about waxing lyrical about the benefits of omitting the AA filter; the Kodak is a perfect example. At 100% view it has the weirdest looking pixels of any camera. Aliasing artifacts all over the place, a peculiar water colour painterly effect and other digital nasties. You can make nice prints from a Kodak but the lack of AA filter has severely damaged the pixel quality. In fact, I now find mine unusable. Images from the 5D are vastly superior.

docmaas wrote:
Good point Derek. Actually I think the superior inherent
separability of pixels in the Sigma is due more to the absence of the
AA filter than the 3 layer chip. Despite the interpolation of bayer
I think you're right that since each position should be just as well
separated in the bayer as in the foveon, and if there are more of
them the finer granularity should lead to better demarcation between
objects. However, the AA filter blurs the separation of pixels on
the bayer filter and that is what really gives them the fuzziness
rather than the interpolation per se. that's why the kodak looks
more like a foveon image or leica image than do other bayer images
with aa filters. Likewise the new Sony A900 is producing, according
to many, inherently sharper images than its competitors and again the
credit is going to a "lighter" Aa filter.

Now that FF sensors have reached far enough to satisfy modern lens
resolution the next improvements will be on making those FF sensors
better in other ways than just more mpix and I suspect that AA
filters will become more sophisticated and perhaps even removable in
the near future on DSLRs as they already are on MF sensors/cameras.

Mike

derek324 wrote:
Some creatures are simpler and even easier to confuse. I tried, and
tried, and tried again, with exactly opposite conclusion.

BTW, as an electronics engineer I would always argue that better
'registration of data' should be expected from sensors placed next to
each other, as opposed to 3 layers of sensors underneath each other,
no metter what the sensors are supposed to 'catch'. I just do not
believe in technology being able to offer total transparency between
the layers, zero reflections, and lack of interference... but this is
of course flogging a dead horse and was discussed many, many times.

Just shows that different perspectives do exists

richard stone wrote:

Simple creature that I am, and as I often say, easily confused, when
I look at these tests, time and time again, the Foveon images look
crisper and clearer to me. No blurring. Each photo-site is
reporting, spatially, exactly what it registered.

...

-- hide signature --
DMillier Forum Pro • Posts: 21,163
Re: my take

The aliasing on the cable near the flywheel is also very clear in this shot.

As Lin pointed out with special software one can reduce this to near invisibility but having to use special software to overcome image defects in not ideal: it forces one down a particular workflow and that limits choice and flexibilty.

Better would be not to have the problem inherent in the file. And aliasing is a normal digital phenomenon that can be treated with am AA filter.

I still think sacrificing the last 5% of sharpness (that is only visible in un-resized 100% screen view anyway) would be a worthwhile trade off to get rid of those damned artifacts at source. Lead balloon suggestion around here of course.

oluv wrote:

many in this thread claim the SD14 to be able to resolve more details
despite of having lower resolution. it indeed resolves an impressive
amount of details for a 4.7 megapixel sensor and it compares very
well to the 50D, but by no means it can resolve more details than the
50D. i have looked at various parts of the image, but i have not
found a single part where the SD14 showed more details than the 50D
file.

you can clearly see that the 50D is able to resolve the ropes on the
very top much better, i can count the exact amount of wire-meshes on
the top of the ladder in the 50D file, in the SD14 file it is still
recognizable as wire mesh, but the are not resolved as single wires
anymore. and also look at the bird sitting on the top of the handrail
is just not resolved in the SD14 file anymore

i attached an image. the 50D file was converted with raw therapee and
sharpened very slightly to compensate for bayer interpolation. the
SD14 was converted with the latest ACR and not sharpened (i do not
have SPP). i didn't want to try any upscaling or downscaling as the
scaled image would always have a disadvantage over the unscaled one.

-- hide signature --
miketuthill Veteran Member • Posts: 3,228
Re: Inferiority Complex

That's very true however it just seems so pointless to me after it's been done so many times before. Kinda like hitting oneself in the head with a hammer continuously and wondering why you have a headache.

DMillier wrote:
No one forces anyone to participate!

miketuthill wrote:
It's fun the first 656 times. By the 657th it starts to get a little
tired.

mike earussi wrote:

Comparisons are fun, that's why we do them.

 miketuthill's gear list:miketuthill's gear list
Sony a77 II Sony a7 II Sony DT 16-105mm F3.5-5.6 Sony DT 18-70mm F3.5-5.6 Sony 75-300mm F4.5-5.6 +14 more
DMillier Forum Pro • Posts: 21,163
Re: Inferiority Complex

It's interesting and absorbing for those who find it interesting and absorbing and painful for those who don't. If you find it painful the rather obvious solution is not to read the threads...

miketuthill wrote:
That's very true however it just seems so pointless to me after it's
been done so many times before. Kinda like hitting oneself in the
head with a hammer continuously and wondering why you have a headache.

DMillier wrote:
No one forces anyone to participate!

miketuthill wrote:
It's fun the first 656 times. By the 657th it starts to get a little
tired.

mike earussi wrote:

Comparisons are fun, that's why we do them.

-- hide signature --
LMC54 Forum Pro • Posts: 11,808
Re: Inferiority Complex

I feel kind of insulted here. I definitely didn't buy because of price nor did I buy to feel inferior. I can buy any camera in the world I want. Actually I paid 1500.00 for a mediocre body. I bought it because I want to use it. I actually have Sigma logos plastered all over my gear and proud of it. It beats the pants off my superior Canons. What can I say, I like how the Foveon per pixel is put together, basically like the structure of how an image is put together. Trust me there is a texture to a bayer image. But who cares just go out and shoot, I'm not here to argue which is better, I'll letl my work speak to my customers.

miketuthill wrote:

All these bayer vs. foveon comparisons over the years have lead me to
believe that some foveon shooters suffer from inferiority complexes
with regards to their dslr choice much as the Nikon vs. Canon vs.
Olympus vs. Sony vs. Pentax....well maybe not Pentax as nobody
bothers to compare to Pentax.....comparators (is that a word???)
exhibit.

IMO, both shots look ok and both would stand up to a normal sized
printing and only a pixel peeper would give a cats ass about the
differences.
--
http://www.pbase.com/miketuthill

Erik Magnuson Forum Pro • Posts: 12,247
Re: Too many sharpening artifacts

Lin Evans wrote:

They both have too many sharpening artifacts including the Canon DPP
conversion. Even the SPP large has bad black sharpening halo's around
the right side of light upright supports.

Sure. My crops were already 200% (nearest neighbor) and then you blew them up another 200%. You could do a better job sharpening for pixel peeping, I was using the in-the-box tools at pretty much default settings for both cameras.

The Photozoom spline interpolation definitely has bad sharpening
artifacts, but I don't believe these are a result of the Photozoom
interpolation but rather that the original was oversharpened.

I didn't say as to why - I was just noting the difference of opinion on what looks good.

-- hide signature --

Erik

miketuthill Veteran Member • Posts: 3,228
Re: Inferiority Complex

No insults intended Larry..if you read my original again you'll notice that I'm lumping all obsessive pixel peeping comparators, regardless of brand affiliation, into the same barrel. As Lin noted maybe I should say "superiority complex":-) I don't dispute their right to continue this fruitless exercise however I maintain my right to make personal observations on the practise.

LMC54 wrote:
I feel kind of insulted here. I definitely didn't buy because of
price nor did I buy to feel inferior. I can buy any camera in the
world I want. Actually I paid 1500.00 for a mediocre body. I bought
it because I want to use it. I actually have Sigma logos plastered
all over my gear and proud of it. It beats the pants off my superior
Canons. What can I say, I like how the Foveon per pixel is put
together, basically like the structure of how an image is put
together. Trust me there is a texture to a bayer image. But who
cares just go out and shoot, I'm not here to argue which is better,
I'll letl my work speak to my customers.

miketuthill wrote:

All these bayer vs. foveon comparisons over the years have lead me to
believe that some foveon shooters suffer from inferiority complexes
with regards to their dslr choice much as the Nikon vs. Canon vs.
Olympus vs. Sony vs. Pentax....well maybe not Pentax as nobody
bothers to compare to Pentax.....comparators (is that a word???)
exhibit.

IMO, both shots look ok and both would stand up to a normal sized
printing and only a pixel peeper would give a cats ass about the
differences.
--
http://www.pbase.com/miketuthill

 miketuthill's gear list:miketuthill's gear list
Sony a77 II Sony a7 II Sony DT 16-105mm F3.5-5.6 Sony DT 18-70mm F3.5-5.6 Sony 75-300mm F4.5-5.6 +14 more
Lin Evans
Lin Evans Forum Pro • Posts: 17,609
Re: Resolution versus detail......

The crops were from the OP's captures. I don't have a RAW converter which will handle the 50D RAW files. No smoothing or any manipulation other than as described was done. Obviously, different results may be seen by different converters, but detail loss is still evident as indicated by Erik in other areas of the image. No converter or sensor is perfect but there are differences which are consistently found and the "mush" at Nyquist is one. This is most likely due to the presence of an AA filter and not to CFA design.

Lin

oluv Senior Member • Posts: 1,973
Re: Resolution versus detail......

indeed the AA filter in the 50D seems to be stronger then in the 40D for example. no idea why canon decided to have it that way, but i prefer some artefacts over blurred details to tell the truth.

 oluv's gear list:oluv's gear list
Panasonic Lumix DMC-GM1 Panasonic Lumix G Vario 7-14mm F4 ASPH Panasonic Lumix G 20mm F1.7 ASPH Panasonic Leica Summilux DG 25mm F1.4 Olympus M.Zuiko Digital 45mm F1.8 +1 more
Lin Evans
Lin Evans Forum Pro • Posts: 17,609
Re: Too many sharpening artifacts

Yes, enlarged simply to show that there are too many sharpening artifacts in all the crops.

Neither did I (say why) - just pointing out that sharpening before enlarging causes issues which can't easily be dealt with in post processing.

Lin

Erik Magnuson wrote:

snip

Sure. My crops were already 200% (nearest neighbor) and then you blew
them up another 200%. You could do a better job sharpening for pixel
peeping, I was using the in-the-box tools at pretty much default
settings for both cameras.

snip

I didn't say as to why - I was just noting the difference of opinion
on what looks good.

-- hide signature --

Erik

LMC54 Forum Pro • Posts: 11,808
Re: Inferiority Complex

Fruitless or not it matters to me what I use. Sorry for my superiority complex.
Maybe it's time for you to get a Leica.

miketuthill wrote:
No insults intended Larry..if you read my original again you'll
notice that I'm lumping all obsessive pixel peeping comparators,
regardless of brand affiliation, into the same barrel. As Lin noted
maybe I should say "superiority complex":-) I don't dispute their
right to continue this fruitless exercise however I maintain my right
to make personal observations on the practise.

LMC54 wrote:
I feel kind of insulted here. I definitely didn't buy because of
price nor did I buy to feel inferior. I can buy any camera in the
world I want. Actually I paid 1500.00 for a mediocre body. I bought
it because I want to use it. I actually have Sigma logos plastered
all over my gear and proud of it. It beats the pants off my superior
Canons. What can I say, I like how the Foveon per pixel is put
together, basically like the structure of how an image is put
together. Trust me there is a texture to a bayer image. But who
cares just go out and shoot, I'm not here to argue which is better,
I'll letl my work speak to my customers.

miketuthill wrote:

All these bayer vs. foveon comparisons over the years have lead me to
believe that some foveon shooters suffer from inferiority complexes
with regards to their dslr choice much as the Nikon vs. Canon vs.
Olympus vs. Sony vs. Pentax....well maybe not Pentax as nobody
bothers to compare to Pentax.....comparators (is that a word???)
exhibit.

IMO, both shots look ok and both would stand up to a normal sized
printing and only a pixel peeper would give a cats ass about the
differences.
--
http://www.pbase.com/miketuthill

miketuthill Veteran Member • Posts: 3,228
Re: Inferiority Complex

See your email Larry.

LMC54 wrote:
Fruitless or not it matters to me what I use. Sorry for my
superiority complex.
Maybe it's time for you to get a Leica.

miketuthill wrote:
No insults intended Larry..if you read my original again you'll
notice that I'm lumping all obsessive pixel peeping comparators,
regardless of brand affiliation, into the same barrel. As Lin noted
maybe I should say "superiority complex":-) I don't dispute their
right to continue this fruitless exercise however I maintain my right
to make personal observations on the practise.

LMC54 wrote:
I feel kind of insulted here. I definitely didn't buy because of
price nor did I buy to feel inferior. I can buy any camera in the
world I want. Actually I paid 1500.00 for a mediocre body. I bought
it because I want to use it. I actually have Sigma logos plastered
all over my gear and proud of it. It beats the pants off my superior
Canons. What can I say, I like how the Foveon per pixel is put
together, basically like the structure of how an image is put
together. Trust me there is a texture to a bayer image. But who
cares just go out and shoot, I'm not here to argue which is better,
I'll letl my work speak to my customers.

miketuthill wrote:

All these bayer vs. foveon comparisons over the years have lead me to
believe that some foveon shooters suffer from inferiority complexes
with regards to their dslr choice much as the Nikon vs. Canon vs.
Olympus vs. Sony vs. Pentax....well maybe not Pentax as nobody
bothers to compare to Pentax.....comparators (is that a word???)
exhibit.

IMO, both shots look ok and both would stand up to a normal sized
printing and only a pixel peeper would give a cats ass about the
differences.
--
http://www.pbase.com/miketuthill

 miketuthill's gear list:miketuthill's gear list
Sony a77 II Sony a7 II Sony DT 16-105mm F3.5-5.6 Sony DT 18-70mm F3.5-5.6 Sony 75-300mm F4.5-5.6 +14 more
Keyboard shortcuts:
FForum MMy threads