Ad
Ad
Ad
Pages: « 1 ... 8 9 [10] 11 »   Bottom of Page
Print
Author Topic: Sigma SD1 review  (Read 44008 times)
Ray
Sr. Member
****
Offline Offline

Posts: 8875


« Reply #180 on: July 29, 2011, 01:03:37 AM »
ReplyReply

I suppose a degree at one end might be called "micro-colorblind?"

Don't be silly, Eric.  Grin

One great advantage of the computer age for the photographer is that one can enlarge any small detail in an image to whatever size one wants.

Here's a random crop of 4 pixels from one of my images. It's been significantly enlarged in Photoshop using nearest neighbour interpolation. It's come up quite well; worthy of any Foveon sensor.

Following is a crop of the same four pixels, but interpolated using bicubic smoother.  Oops! No amount of sharpening could fix that.

Now I'm tempted to suggest that the differences between these two versions is conceptually similar to the differences between the Bayer array and the Foveon sensor, but I stress the word conceptually, and I admit I might be drawing a long bow.  Grin

Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7309


WWW
« Reply #181 on: July 29, 2011, 02:58:42 AM »
ReplyReply

Hi,

To my understanding "the funnel" represents MTF and MTF is shrinking with distance due to target having higher and higher frequency.

The RED goes much further to the right than the Alexa, is this due to higher sensor resolution, different OLP filtering or both? What would the plot look like without OLP?

Best regards
Erik


To my mind, "Clarity" or other such sliders enhances large area contrast - it does nothing for small scale contrast. If you look at what the effect is - unsharp mask with a very very large radius and low amount, you can see it's a large area effect. It's enhancing the low frequency MTF, not the highest frequency MTF which traditional use of unsharp mask alters.

What it comes down to is we need an MTF plot rather than talking a single number for resolution. We need to be able to see how much contrast we get for all frequencies represented in the image and also to be able to see how much aliasing occurs and how much that is contributing to the MTF.

http://reduser.net/forum/showthread.php?58227-Leica-Lens-Resolution-test...&p=759177&viewfull=1#post759177 are not exactly MTF plots, but they're related and show pretty clearly what's going on.

Graeme
Logged

hjulenissen
Sr. Member
****
Offline Offline

Posts: 1675


« Reply #182 on: July 29, 2011, 04:09:45 AM »
ReplyReply

With absolutely precise accuracy. One pixel has a value of R=120, G=180, B=240, and the adjacent pixel also has a value of R=120, G=180, B=240. Result? Microcontrast zero between these two pixels.

However, if the adjacent pixel has a value of R=320, G=140, B=89, then we have a specific and clearly defined degree of microcontrast betwee the two pixels. What's your probledm?  Grin
*According to your definition, sensor noise improves micro contrast?
*Many (most?) scenes does not naturally contain infinite contrast transitions. What scenes do you suggest using to estimate the camera "micro contrast" capability?
*Most cameras would (according to your definition) contain millions of micro contrast readings for a single image. How do you propose to manage that data set?
*Are these sRGB pixels? Camera raw pixels?
*Does my former 8megapixel 350D have more microcontrast than my 18 megapixel 7D since it commonly have larger pixel-to-pixel differences?
*It seems to me that your definition of micro-contrast is the same as a discrete 1st-order gradient?
*It seems to me that your definition of micro-contrast is the same as the MTF at the nyquist frequency?
*If you move your camera 1/2 pixel, many cameras will get different micro-contrast using your measure (due to aliasing), especially olpf-less cameras like the Sigmas. Which reading should be chosen?

-h
« Last Edit: July 29, 2011, 04:14:13 AM by hjulenissen » Logged
Dave Millier
Full Member
***
Offline Offline

Posts: 118


WWW
« Reply #183 on: July 29, 2011, 06:43:01 AM »
ReplyReply

Ouch!

It seems that Ray's argument is of the type characterised by Richard Dawkins as "argument from personal incredulity"...



*According to your definition, sensor noise improves micro contrast?
*Many (most?) scenes does not naturally contain infinite contrast transitions. What scenes do you suggest using to estimate the camera "micro contrast" capability?
*Most cameras would (according to your definition) contain millions of micro contrast readings for a single image. How do you propose to manage that data set?
*Are these sRGB pixels? Camera raw pixels?
*Does my former 8megapixel 350D have more microcontrast than my 18 megapixel 7D since it commonly have larger pixel-to-pixel differences?
*It seems to me that your definition of micro-contrast is the same as a discrete 1st-order gradient?
*It seems to me that your definition of micro-contrast is the same as the MTF at the nyquist frequency?
*If you move your camera 1/2 pixel, many cameras will get different micro-contrast using your measure (due to aliasing), especially olpf-less cameras like the Sigmas. Which reading should be chosen?

-h
Logged

My website and photo galleries: http://www.whisperingcat.co.uk/
Graeme Nattress
Sr. Member
****
Offline Offline

Posts: 582



WWW
« Reply #184 on: July 29, 2011, 07:44:40 AM »
ReplyReply

Ray, the problem with your model is that you're ignoring noise. Noise will either manifest as thus - fake "micro-contrast" on flat regions that don't have any, or when necessarily noise reduced you'll loose any differentiation in the chroma.

If MTF at sampling limits is micro-contrast then almost by definition it's another description of aliasing.

Erik, plots are on a linear sine zone plate. It's not due to lens - same lens on both and same aperture / focus. It's three things - sensor resolution (5120 across v 2880 across recorded area), OLPF and image processing (demosaic). It's not just that RED goes further to the right, it's the area that is proportional to perceived sharpness, and how the MTF stays higher even at the lower frequencies - ie the MTF50 figure is better both absolute and in proportion to limiting resolution. Without OLPF both resolution and aliasing increase. This is not wise on a motion camera and already we see too much aliasing on the Arri so weakening or removing the OLPF there would be very unwise.

Graeme
Logged

www.nattress.com - Plugins for Final Cut Pro and Color
www.red.com - Digital Cinema Cameras
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7309


WWW
« Reply #185 on: July 29, 2011, 04:01:08 PM »
ReplyReply

Hi,

I presume that the ultimate solution to achieving maximum resolution with excellent micro contrast is to reduce pixel size to perhaps 3 microns so lens aberrations, diffraction light diffusion in silicon and defocusing will limit MTF to reasonable values at Nyquist?

That would of course reduce DR significantly, about one stop if the image is downsampled to 6 micron pixels?

Best regards
Erik

Ray, the problem with your model is that you're ignoring noise. Noise will either manifest as thus - fake "micro-contrast" on flat regions that don't have any, or when necessarily noise reduced you'll loose any differentiation in the chroma.

If MTF at sampling limits is micro-contrast then almost by definition it's another description of aliasing.


Graeme
Logged

Graeme Nattress
Sr. Member
****
Offline Offline

Posts: 582



WWW
« Reply #186 on: July 29, 2011, 05:22:19 PM »
ReplyReply

Smaller pixels will get better as the technology improves. The other thing is larger sensors with the same size pixel, and that leads to it's own issues. There's no easy answer other than to keep moving forwards as best as we can!

Graeme
Logged

www.nattress.com - Plugins for Final Cut Pro and Color
www.red.com - Digital Cinema Cameras
Ray
Sr. Member
****
Offline Offline

Posts: 8875


« Reply #187 on: July 29, 2011, 06:42:04 PM »
ReplyReply

Ray, the problem with your model is that you're ignoring noise. Noise will either manifest as thus - fake "micro-contrast" on flat regions that don't have any, or when necessarily noise reduced you'll loose any differentiation in the chroma.

Graeme

I never ignore noise Graeme. I'm very sensitive to it. I recall when using my first DSLR, the Canon D60, I was reluctant to use even ISO 400 because noise became clearly apparent. ISO 800 was pretty awful. Even with my latest camera, the Nikon D7000, noise is still apparent at ISO 800, but a vast improvement on the D60 of course.

The reason I might appear to be ignoring the noise characteristics of the SD1 is I simply don't have any reliable information on the subject. DXO Mark have not published their tests yet.

But you are quite right, if noise proves to be an issue at reasonably low ISOs, compared with the latest Bayer type sensors, then that would be another reason not to buy into the system. There has to be a compelling reason for me to buy new equipment. So far, I can think of only one compelling reason why I would even consider getting the SD1, and that is its higher 'effective' resolution compared with existing cropped-format DSLRs.

However, that factor by itself would not necessarily be sufficient reason for me if there are other negative factors such as more noticeable noise at ISO 800 and 1600, compared with Nikon and Canon, and the lack of image stabilisation in certain Sigma lenses, particularly the 800/F5.6.

Logged
Graeme Nattress
Sr. Member
****
Offline Offline

Posts: 582



WWW
« Reply #188 on: July 29, 2011, 07:06:16 PM »
ReplyReply

Yes Ray, there's an awful lot of factors to take into account! That's what makes this all interesting, and doubly so when Sigma try an alternative path - there are tradeoffs and we've got to know where they are to make a good reasoned decision.

Graeme
Logged

www.nattress.com - Plugins for Final Cut Pro and Color
www.red.com - Digital Cinema Cameras
Ray
Sr. Member
****
Offline Offline

Posts: 8875


« Reply #189 on: July 29, 2011, 07:27:07 PM »
ReplyReply

*According to your definition, sensor noise improves micro contrast?
*Many (most?) scenes does not naturally contain infinite contrast transitions. What scenes do you suggest using to estimate the camera "micro contrast" capability?
*Most cameras would (according to your definition) contain millions of micro contrast readings for a single image. How do you propose to manage that data set?
*Are these sRGB pixels? Camera raw pixels?
*Does my former 8megapixel 350D have more microcontrast than my 18 megapixel 7D since it commonly have larger pixel-to-pixel differences?
*It seems to me that your definition of micro-contrast is the same as a discrete 1st-order gradient?
*It seems to me that your definition of micro-contrast is the same as the MTF at the nyquist frequency?
*If you move your camera 1/2 pixel, many cameras will get different micro-contrast using your measure (due to aliasing), especially olpf-less cameras like the Sigmas. Which reading should be chosen?

-h

h,
There are only two instruments I use to test image qualities and charcteristics; my two eyes.

That said, I find it useful to have a set of technical descriptions which tend to closely match, or are consistent with, the impressions I get using my own eyes. It's why I'm impressed with the test results from DXO Mark. They seem consistent with what I see, and I have taken the trouble to make comparative tests of my own cameras to see how they compare with the comparative results at DXO Mark.

To put it another way, if I compare images from two cameras, of the same scene, and the images from one of the cameras look more detailed than the images from the other camera, yet don't appear to have more noise, it may be of only academic interest to be informed that the more detailed images actually do have more noise.

In other words, if the noise is so well disguised that it's not apparent or visible through those two instruments, my eyes, and if the images still appear to be more detailed despite supposedly containing more noise, then I'see no reason to be unduly worried or concerned about the issue.

However, if I were using the camera as a scientific instrument to gather precise data for some research purposes, I would likely be very concerned about the issue of aliasing and the possibility of receiving false data or incorrect data which has resulted from noise or attempts to remove it.
Logged
hjulenissen
Sr. Member
****
Offline Offline

Posts: 1675


« Reply #190 on: July 30, 2011, 05:18:15 AM »
ReplyReply

I feel like we are argueing in a loop here.
h,
There are only two instruments I use to test image qualities and charcteristics; my two eyes.
Does this mean that your eyes return discrete values for r,g and b? Seems to me that two posts ago you suggested a method that was not based on using your eyes at all:
Quote
With absolutely precise accuracy. One pixel has a value of R=120, G=180, B=240, and the adjacent pixel also has a value of R=120, G=180, B=240. Result? Microcontrast zero between these two pixels.

However, if the adjacent pixel has a value of R=320, G=140, B=89, then we have a specific and clearly defined degree of microcontrast betwee the two pixels
Quote
That said, I find it useful to have a set of technical descriptions which tend to closely match, or are consistent with, the impressions I get using my own eyes. It's why I'm impressed with the test results from DXO Mark. They seem consistent with what I see, and I have taken the trouble to make comparative tests of my own cameras to see how they compare with the comparative results at DXO Mark.

To put it another way, if I compare images from two cameras, of the same scene, and the images from one of the cameras look more detailed than the images from the other camera, yet don't appear to have more noise, it may be of only academic interest to be informed that the more detailed images actually do have more noise.

In other words, if the noise is so well disguised that it's not apparent or visible through those two instruments, my eyes, and if the images still appear to be more detailed despite supposedly containing more noise, then I'see no reason to be unduly worried or concerned about the issue.

However, if I were using the camera as a scientific instrument to gather precise data for some research purposes, I would likely be very concerned about the issue of aliasing and the possibility of receiving false data or incorrect data which has resulted from noise or attempts to remove it.
Using practical, hands-on subjective impressions is all fine and good. After all, that is what guide me and you through 99% of the choices that we face in our everyday life. I am sceptical about coining the term "micro contrast" based only on subjective impressions because you and Michael may not agree what it is, and a 3rd person might have another feeling about what it is. Using the word "contrast" seems to suggest that it has a physical meaning - bound to cause confusion.

If you really only care about subjective impressions, then technical and quasi-technical stuff should be irrelevant to you. So why care about micro-contrast or MTF50 at all? If you, like me, think that great art can be the result of subjective _and_ objective components, then we are back to the original discussion?

-h
« Last Edit: July 30, 2011, 05:22:43 AM by hjulenissen » Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3553


« Reply #191 on: July 30, 2011, 10:29:18 AM »
ReplyReply

I wish people would shoot better charts on the Foveon and then I'd have better numbers to go on there - so instead I just give it a 100% luma resolution factor (usually reckon Bayer CFA with OLPF around 78%).

I've made available a target for download and print for the seriously interested amongst us at:
http://www.openphotographyforums.com/forums/showthread.php?t=13217

I always hesitate to divert people to other websites (it's not nice to our host), but in this case I wanted to avoid posting the whole story again (hope Michael forgives me). Discussions about those targets, and the results people get, can take place here (again sorry Michael Wink ).

Quote
What I'm keen for photographers to understand is what measurable aspects of photography correlate with what they see visually. I think that's useful and powerful for them to know so that they can understand why they like what they like and how they can use that to their visual advantage.

Same with me. There is a difference between looking at the aliasing patterns of almost parallel (hyperbolic) resolution test patterns such as those on DPreview, and targets that test resolution/aliasing at many angles. My target comes closer to looking at a brick wall or road at an angle, and that won't result in an aliasing pattern that mimics the original pattern ...

Cheers,
Bart
« Last Edit: August 04, 2011, 04:41:03 AM by BartvanderWolf » Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3553


« Reply #192 on: July 30, 2011, 10:56:04 AM »
ReplyReply

I am not trying to be difficult, but I think that you guys are inventing a vague pseudo-term for something that allready has a better name. How do you measure micro-contrast?

We might take a look at how the ISO resolution determination specifies "limiting resolution". It's the point on the MTF curve where 10% response remains. Apparently that corresponds to human visual resolution limits as well. The only difficulty is that many lens/sensor combinations, even those with OLPFs, have no difficulty in reaching the Nyquist limit with still 10-20% response. Therefore the ISO suggests 10% MTF response or the Nyquist frequency, whichever is reached first, as the limiting resolution.

At those micro detail levels, the contrast reproduction of our capture system is suffering considerable loss, which makes the term Micro-contrast a very apt description of the MTF curve of the capture system falling to a level where human vision also has its limit in resolving capabilities.

Cheers,
Bart
Logged
crames
Full Member
***
Offline Offline

Posts: 210


WWW
« Reply #193 on: July 30, 2011, 07:11:29 PM »
ReplyReply

I'm made available a target for download and print for the seriously interested amongst us at:
http://www.openphotographyforums.com/forums/showthread.php?t=13217

Hi Bart,

Are you familiar with the method of Lionel Baker, called OPW or Optimum Print Width, (another description here)? He uses a 36 segment "sector star" instead of the sine star. He determines the Optimum Print Width from the blur-circle size, Optimum Print Width being the largest image that can be printed with full detail visible to the eye at arms length. It's a neat way to compare cameras because it avoids the need to up-sample or down-sample to a common print size.

I made a 36 sector sine star (available here, large version here), which can be turned into the "sector star" with Photoshop's Threshhold tool, but I don't think there is much difference in the results between the sine and the sector versions.

Regards,
« Last Edit: July 30, 2011, 07:21:22 PM by crames » Logged

Cliff
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3553


« Reply #194 on: July 30, 2011, 08:50:35 PM »
ReplyReply

Hi Bart,

Are you familiar with the method of Lionel Baker, called OPW or Optimum Print Width, (another description here)? He uses a 36 segment "sector star" instead of the sine star. He determines the Optimum Print Width from the blur-circle size, Optimum Print Width being the largest image that can be printed with full detail visible to the eye at arms length. It's a neat way to compare cameras because it avoids the need to up-sample or down-sample to a common print size.

Hi Cliff,

Thanks for that link. I wasn't aware of that document, which uses a similar technique. The drawback of a 36 segment star is that it requires a large shooting distance to produce some unresolved detail at a diameter that's large/accurate enough for numerical evaluation. But for visual use, even when it involves a computer display, it could be adequate.

Quote
I made a 36 sector sine star (available here, large version here), which can be turned into the "sector star" with Photoshop's Threshhold tool, but I don't think there is much difference in the results between the sine and the sector versions.

It won't make much difference with a dithered print method or when used visually. However, when shooting it with a discrete sampling device, a sine target of sorts is always preferrable to bi-tonal versions. One wants to avoid the presence of higher spatial frequencies (sharp edges) interfering with the spatial frequencies under investigation. That's why I already introduced a 60 cycle sinusoidal version of a Siemens star on Usenet back in 2003, when printers were not as good yet as they are today.

Cheers,
Bart
« Last Edit: July 31, 2011, 05:11:34 AM by BartvanderWolf » Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8875


« Reply #195 on: July 30, 2011, 09:04:16 PM »
ReplyReply

I feel like we are argueing in a loop here.Does this mean that your eyes return discrete values for r,g and b? Seems to me that two posts ago you suggested a method that was not based on using your eyes at all:Using practical, hands-on subjective impressions is all fine and good. After all, that is what guide me and you through 99% of the choices that we face in our everyday life. I am sceptical about coining the term "micro contrast" based only on subjective impressions because you and Michael may not agree what it is, and a 3rd person might have another feeling about what it is. Using the word "contrast" seems to suggest that it has a physical meaning - bound to cause confusion.
-h

h,
I think Graeme expressed it very well a few posts ago, as follows:

Quote
What I'm keen for photographers to understand is what measurable aspects of photography correlate with what they see visually. I think that's useful and powerful for them to know so that they can understand why they like what they like and how they can use that to their visual advantage.

Now we both know that our eyes alone cannot determine precise values of RGB in a shade without the aid of another tool such as Photoshop.

What's important visually is the detection of differences between shades, and that also applies to many other technical parameters of image characteristics, such as the MTF response of a lens.

If the difference is not great enough for it to be visually significant, then perhaps it doesn't matter from the perspective of the person producing an image.

For example, I might decide to buy a particular lens because it is claimed to have a higher MTF response than another lens at a particular fequency. It is the degree of difference, visually apparent in an image, resulting from the differences in MTF response of different models of lenses, also influenced by other considerations such as price, that helps me make a decision. If the degree of difference is so small visually, whether in absolute terms or because such differences are obscured by other factors, then I hope I would have the sense not to be influenced by such a difference.

Quote
If you really only care about subjective impressions, then technical and quasi-technical stuff should be irrelevant to you. So why care about micro-contrast or MTF50 at all? If you, like me, think that great art can be the result of subjective _and_ objective components, then we are back to the original discussion?

I've never written that I only care about subjective impressions. I don't know how you could have drawn that conclusion. I like the term mocrocontrast because it is so apt, so intuitive to understand and so fundamental to the modern digital process of image making. The word contrast is fundamental to all image making, whether photography or painting. No contrast, no picture. A contrastless picture cannot exist. It's an oxymoron, unless you consider a totally uniform area of plain shade, devoid of any detail, a modern work of art perhaps.  Grin

When addressing matters of contrast in digital images at the pixel level, or at the level of small groups of pixels, the term micro seems very appropriate to me, and meaningful. The term microcontrast is as precise as it needs to be in communicating a visual effect.

However, we should also be aware that there are certain more global visual phenomena that seem to defy the technical descriptions at the micro level.

I'm no expert in this field, but the following image illustrates very well how the context surrounding an area with a specific and precisely defined RGB value, can create the impression, visually, that the RGB value of another area in a different context has a different RGB value, when in fact it is the same.

It's clear to me that one of those orange circles is a different shade to the other. It's more brown than orange. Not only that; the grey square that contains the brown circle is very noticeably darker than the grey square that contains the paler orange circle.

Now I can't help wondering if there are any viewers who see both orange circles as being the same shade, and see the grey squares that contain each of them as being the same shade, because they actually are the same shade technically.

Both orange circles have the same RGB value, and both squares also have the same RGB values. But please, no need to rush off to see a psychiatrist.  Grin




Logged
joofa
Sr. Member
****
Offline Offline

Posts: 486



« Reply #196 on: July 30, 2011, 09:38:07 PM »
ReplyReply

I think Ray has made it clear that micro-contrast means a variation in the local neighborhood of a pixel. And, that is a valid notion to consider, IMHO. Such analyses are common in signal processing, and I think one can live with the fact that others can have a different name for the same/similar phenomenon that one may have learnt in their domain under a different name or context. The way I see it is that it is important to understand what underlying phenomenon a term such as micro-contrast entails than arguing on name of the phenomenon itself.

Joofa
Logged

Joofa
http://www.djjoofa.com
Download Photoshop and After Effects plugins
peterzpicts
Newbie
*
Offline Offline

Posts: 12


« Reply #197 on: August 01, 2011, 11:55:11 PM »
ReplyReply

Now I know why real photographers have their hair set on fire by some of the nauseating technical discussions that a thread can evolve into. I am a nerd at heart but sometimes its best to get out of the technical arena and let the eyeballs make the call.

Lets just call this "why I continue to inflict the pain using of Sigma cameras"
These were taken during a portrait sitting shared from Smugmug resized to X3 by smuggie to make things "fair". I am sure their are a ton of technical flaws with this demonstration, I am a hobbiest not a camera reviewer.

Exhibit 1
Portrait taken with SD14 and 18-200DCOS lens. Processed with SPP 5 corrected in iPhoto and bumped "definition" whatever that means but I like it.

Exhibit 2 Same sitting taken with Nikon D90 and Nikkor 35DX 1.8 Incamera Jpeg, I find it hard to beat Nikon's incamera processing for shots like this.


The bottom line for me I like what I see from Sigma better, and why many others continue to bother with their under engineered equipment.
Paul
« Last Edit: August 02, 2011, 05:08:23 AM by peterzpicts » Logged
ilsiu
Newbie
*
Offline Offline

Posts: 7


« Reply #198 on: August 02, 2011, 08:08:20 AM »
ReplyReply

Regarding Paul's examples, my impressions are that the Sigma image is warmer and I think it give the subject more pop, which is more pleasing than the Nikon image.  However, I think it would be pretty straightforward to process the Nikon image to match the warmth in the Sigma image.

I really don't see any appreciable difference in detail or sharpness.  The eyes are quite sharp and hair strands are nicely defined in both photos.

Subjectively, I don't like the welcome sign in the upper left corner of the Sigma image (can be easily removed w/content aware fill), and I like the smile in the Nikon image much better (maybe she's happier and more relaxed when she saw you using the Nikon  Grin).

Both are nice portraits, but I prefer the Nikon for non-technical reasons (good smile).  I have trouble seeing any technical advantages that can't be negated with processing.
Logged
michael
Administrator
Sr. Member
*****
Offline Offline

Posts: 4789



« Reply #199 on: August 02, 2011, 05:12:10 PM »
ReplyReply

The Nikon image simply has a different white balance and brightness. It would take 10 seconds to make it match the Sigma frame if that's what one wanted.

Michael
Logged
Pages: « 1 ... 8 9 [10] 11 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad