Capture 3.5 - Pics of 5MP vs 10MP

Here's what I did...

Each image was opened with Capture 3.5, and then sent to Photoshop 7 -- The first at 5MP and the second at 10MP.

I then resized the 5MP image to the same dimensions as the 10MP image using Photoshop's resampling tool so I could compare Capture's 10MP image output to Photoshop's.

No image sharpening or tweaking was done to either image. Both crops are at 100% magnification.

Here's the resized 5MP image:



Here's the Capture 10MP image:



Capture's 10 MP output is of much better quality in my opinion.

--Abe

http://www.abekleinfeld.com

Sharpening and contrast are enhanced as part of the processing of the 10MP image, but the image quality itself is lower. Check the edges for yourself, esp. high-contrast edges, even in these two images. Look at smaller details as well.

I've done a comparative analysis also, and am noticing blatant interpolation artifacts even over the obvious artifacts present in a "normally"-interpolated 6MP D1x image. While the 10MP output would be fine on matte, canvas or watercolor paper and viewed at a reasonable distance, when viewed close-up (on a monitor or on glossy paper) it has problems.

Ron

PS: if there is interest, I can post a 1:1 crop comparison, but I'm sure there are going to be tons of them around in the next few days.

Abe said:

Here's what I did...

Each image was opened with Capture 3.5, and then sent to Photoshop
7 -- The first at 5MP and the second at 10MP.

I then resized the 5MP image to the same dimensions as the 10MP
image using Photoshop's resampling tool so I could compare
Capture's 10MP image output to Photoshop's.

No image sharpening or tweaking was done to either image. Both
crops are at 100% magnification.

Here's the resized 5MP image:



Here's the Capture 10MP image:



Capture's 10 MP output is of much better quality in my opinion.

--Abe

http://www.abekleinfeld.com

Click to expand...

--
Ron Reznick
http://digital-images.net
http://trapagon.com

I've been using Bibble up until Capture 3, now I use both because I'm still trying to decide which is best.

Ron, I'd like to see and hear your comments on the 10MP image from NC3.5. I've noticed that the 10MP files from Bibble get oversharpened even with the setting at normal. I'm going to try and make a detailed comparrison as soon as possible, but in the meantime, I'm curious to see your example.

--
Regards,
Joe H.

PPA
---------------------------------------
http://www.biggerboatstudios.com

(Sarcasm Included - some assembly required.)

I compared an image shot as NEF using NC 3.5 at the larger size, then enlarged a jpeg (same subject, jpeg made in camera) to same size, about 50% larger, using SI. At 100% looked the same, detail was the same, all interpolation artifacts matched. I'm unimpressed.

Batch processing is much improved, and nice that it works in OS X finally.

NEF has its place, if you're in a tricky situation where you may need to correct the image (my SB unit is flaky, and may try NEF till it's fixed when I have to use it), but if you shoot digital like you're shooting chromes, there's still no difference in final product. No reason to waste the card space, processing time, etc.

--
-T

Just check a sample of my previous shot NEF, the 10MP file definitely produces very impression quality, it seems introducing a bit more sharpening and contrast which result in a bit more noise.

I find a gaussian blur at 0.3 help to keep the over run smoother on the 10MP version at my iso400 sample, it turns out almost as the same as that of the 6MP version.

I also tested the noise reduction on my high iso sample, it works, I find level 3 probably the best for my case at iso400.

GH

Abe said:

Here's what I did...

Each image was opened with Capture 3.5, and then sent to Photoshop
7 -- The first at 5MP and the second at 10MP.

I then resized the 5MP image to the same dimensions as the 10MP
image using Photoshop's resampling tool so I could compare
Capture's 10MP image output to Photoshop's.

No image sharpening or tweaking was done to either image. Both
crops are at 100% magnification.

Here's the resized 5MP image:



Here's the Capture 10MP image:



Capture's 10 MP output is of much better quality in my opinion.

--Abe

http://www.abekleinfeld.com

Click to expand...

--



http://www.digital-life-21.com/forums/
Nikon D1x, Nikon FM3A, Canon S45

That probably its value, but not in terms of quality.

GH

Warning: Don't Read said:

I compared an image shot as NEF using NC 3.5 at the larger size,
then enlarged a jpeg (same subject, jpeg made in camera) to same
size, about 50% larger, using SI. At 100% looked the same, detail
was the same, all interpolation artifacts matched. I'm unimpressed.

Batch processing is much improved, and nice that it works in OS X
finally.

NEF has its place, if you're in a tricky situation where you may
need to correct the image (my SB unit is flaky, and may try NEF
till it's fixed when I have to use it), but if you shoot digital
like you're shooting chromes, there's still no difference in final
product. No reason to waste the card space, processing time, etc.

--
-T

Click to expand...

--



http://www.digital-life-21.com/forums/
Nikon D1x, Nikon FM3A, Canon S45

Joe Hoddinott said:

I've been using Bibble up until Capture 3, now I use both because
I'm still trying to decide which is best.

Ron, I'd like to see and hear your comments on the 10MP image from
NC3.5. I've noticed that the 10MP files from Bibble get
oversharpened even with the setting at normal. I'm going to try
and make a detailed comparrison as soon as possible, but in the
meantime, I'm curious to see your example.

Click to expand...

just ran a preliminary test of 10 Mpix Bibble vs Capture 3.5 from my D1X and the results are conflicting. No doubt Bibble produces more "true", life-like colours than NC 3.5 without any fiddling with colour settings, whilst NC3.5 has slightly better rendition of minute details, in particular in the shadows and lower mid tones. I use sharpening OFF in the camera, which has firmware 5.0 (+ memory upgrade), and don't see any oversharpening side effects either in Bibble or NC 3.5 outputs. Both Bibble and NC3.5 give 6 MPix files from D1X with less detail than the 10 MPix versions.

Bibble still runs much faster because it is optimised for dual-CPU machines which are my preferred configuration. Capture interferes with NikonScan and my LS-8000 scanner so I'm not too keen to implement this software on my win2000 workstation, and of course Capture - unlike Bibble - won't run on my much faster NT boxes either, so I'm in a bind here what to do with it. Presumably I'll keep it on one of my laptops only, unless further testing convinces me NC3.5 is really superior to Bibble.

Ron Reznick said:

Sharpening and contrast are enhanced as part of the processing of
the 10MP image, but the image quality itself is lower. Check the
edges for yourself, esp. high-contrast edges, even in these two
images. Look at smaller details as well.

I've done a comparative analysis also, and am noticing blatant
interpolation artifacts even over the obvious artifacts present in
a "normally"-interpolated 6MP D1x image. While the 10MP output
would be fine on matte, canvas or watercolor paper and viewed at a
reasonable distance, when viewed close-up (on a monitor or on
glossy paper) it has problems.

Click to expand...

I have spent most of the day today playing with 7.7MP vs 9.8MP NEFs from Capture 3.5. I have peered at all areas of several images (both low and high contrast) and I don't see lower image quality on the 9.8 and I haven't seen a single artifact. IMO, I the images, overall, look quite a bit better.

Member said:

Ron

PS: if there is interest, I can post a 1:1 crop comparison, but I'm
sure there are going to be tons of them around in the next few days.

Click to expand...

I would be quite interested in something that shows the difference you see.

Thanks,

-Tom

Joe, in the image below cropped from a set of TIFs converted in NC3.5 and processed identically other than the output to 10MP and 6MP (by opening with the appropriate preference set), you will notice the additional sharpening applied by NC's processing to the 10MP image, the additional moire in the layered feathers, the broken, jagged character of the upper-layer feathers on the left side of the image, and the blatant contrast adjustment applied in the 10MP processing by Capture. There are also noticeable sharpening artifacts on higher-contrast edges.

While the initial impression is that there is more detail, it is artificially-induced detail and the addition of aliasing and sharpening artifacts along with the additional interpolation artifacts are somewhat problematic for me. It is possible that given time I could figure out a method of altering the processing of the 10MP image to reduce the artifacts somewhat (e.g. processing for lower contrast, using no USM, etc.), but considering the fact that there are some other problems with 3.5 that will interfere with my workflow (notably the Set Grey problem in the WB dialog), I think I'm going to pass on this upgrade until they get a few things straightened out.



Ron

Joe Hoddinott said:

I've been using Bibble up until Capture 3, now I use both because
I'm still trying to decide which is best.

Ron, I'd like to see and hear your comments on the 10MP image from
NC3.5. I've noticed that the 10MP files from Bibble get
oversharpened even with the setting at normal. I'm going to try
and make a detailed comparrison as soon as possible, but in the
meantime, I'm curious to see your example.

--
Regards,
Joe H.

PPA
---------------------------------------
http://www.biggerboatstudios.com

(Sarcasm Included - some assembly required.)

Click to expand...

--
Ron Reznick
http://digital-images.net
http://trapagon.com

I find that the 10 MP NEF give quite nice result if sharpening is set to none for the picture before conversion.

GH

Bjorn Rorslett said:

Joe Hoddinott said:

I've been using Bibble up until Capture 3, now I use both because
I'm still trying to decide which is best.

Ron, I'd like to see and hear your comments on the 10MP image from
NC3.5. I've noticed that the 10MP files from Bibble get
oversharpened even with the setting at normal. I'm going to try
and make a detailed comparrison as soon as possible, but in the
meantime, I'm curious to see your example.

Click to expand...

just ran a preliminary test of 10 Mpix Bibble vs Capture 3.5 from
my D1X and the results are conflicting. No doubt Bibble produces
more "true", life-like colours than NC 3.5 without any fiddling
with colour settings, whilst NC3.5 has slightly better rendition of
minute details, in particular in the shadows and lower mid tones. I
use sharpening OFF in the camera, which has firmware 5.0 (+ memory
upgrade), and don't see any oversharpening side effects either in
Bibble or NC 3.5 outputs. Both Bibble and NC3.5 give 6 MPix files
from D1X with less detail than the 10 MPix versions.

Bibble still runs much faster because it is optimised for dual-CPU
machines which are my preferred configuration. Capture interferes
with NikonScan and my LS-8000 scanner so I'm not too keen to
implement this software on my win2000 workstation, and of course
Capture - unlike Bibble - won't run on my much faster NT boxes
either, so I'm in a bind here what to do with it. Presumably I'll
keep it on one of my laptops only, unless further testing convinces
me NC3.5 is really superior to Bibble.

Click to expand...

--



http://www.digital-life-21.com/forums/
Nikon D1x, Nikon FM3A, Canon S45

Ron,

I agree there's no real magic happening here and certainly won't argue with your fine analysis. After all, getting a 10MP image from a 5MP sensor naturally will not produce the same quality as a native 10MP sensor. However, the reason I believe you'd want a 10MP image in the first place is to output a larger image. So the relevant question is whether Capture is better at producing a 13.4 x 8.7 inch, 300 dpi image from a D1X than Photoshop or some other program.

I don't have Bibble or other programs to test, but to my eye Capture does a better job of outputting a 10MP-sized image than Photoshop 7.

--Abe

http://www.abekleinfeld.com

Ron Reznick said:

I've done a comparative analysis also, and am noticing blatant
interpolation artifacts even over the obvious artifacts present in
a "normally"-interpolated 6MP D1x image. While the 10MP output
would be fine on matte, canvas or watercolor paper and viewed at a
reasonable distance, when viewed close-up (on a monitor or on
glossy paper) it has problems.

Ron

PS: if there is interest, I can post a 1:1 crop comparison, but I'm
sure there are going to be tons of them around in the next few days.

Abe said:

Here's what I did...

Each image was opened with Capture 3.5, and then sent to Photoshop
7 -- The first at 5MP and the second at 10MP.

I then resized the 5MP image to the same dimensions as the 10MP
image using Photoshop's resampling tool so I could compare
Capture's 10MP image output to Photoshop's.

No image sharpening or tweaking was done to either image. Both
crops are at 100% magnification.

Here's the resized 5MP image:



Here's the Capture 10MP image:



Capture's 10 MP output is of much better quality in my opinion.

--Abe

http://www.abekleinfeld.com

Click to expand...

--
Ron Reznick
http://digital-images.net
http://trapagon.com

Click to expand...

Hi all,

all this 10MP D1x talk is very interesting. As someone has said, lots of examples are going to get posted. I'd be keen to have the links to some very detail rich shots at 10MP posted if possible. Say a cityscape panorama or something similarly detailed, that would be great.

You can only ask, I guess....)
--
Regards

Andrew McGregor

http://andrewmcgregor.tripod.com

I agree there are times when upsampling is necessary -- there are a number of methods available and you need to decide which works best for you given the character of the output.

Ron

Abe said:

Ron,

I agree there's no real magic happening here and certainly won't
argue with your fine analysis. After all, getting a 10MP image from
a 5MP sensor naturally will not produce the same quality as a
native 10MP sensor. However, the reason I believe you'd want a 10MP
image in the first place is to output a larger image. So the
relevant question is whether Capture is better at producing a 13.4
x 8.7 inch, 300 dpi image from a D1X than Photoshop or some other
program.

I don't have Bibble or other programs to test, but to my eye
Capture does a better job of outputting a 10MP-sized image than
Photoshop 7.

--Abe

http://www.abekleinfeld.com

Click to expand...

--
Ron Reznick
http://digital-images.net
http://trapagon.com

Ron Reznick said:

Joe, in the image below cropped from a set of TIFs converted in
NC3.5 and processed identically other than the output to 10MP and
6MP (by opening with the appropriate preference set), you will
notice the additional sharpening applied by NC's processing to the
10MP image

Click to expand...

I think using TIF's is not the proper file format. You have to use NEF.

Within another thread at this forum, showing a landscape, do find the 10MP pictures better than a interpolated 6 MP afterwards within Photoshop.

--
Leon Obers

Leon, I shoot NEF only. I converted to TIF in Capture 3.5.

Ron

Leon Obers said:

Ron Reznick said:

Joe, in the image below cropped from a set of TIFs converted in
NC3.5 and processed identically other than the output to 10MP and
6MP (by opening with the appropriate preference set), you will
notice the additional sharpening applied by NC's processing to the
10MP image

Click to expand...

I think using TIF's is not the proper file format. You have to use
NEF.
Within another thread at this forum, showing a landscape, do find
the 10MP pictures better than a interpolated 6 MP afterwards within
Photoshop.

--
Leon Obers

Click to expand...

--
Ron Reznick
http://digital-images.net
http://trapagon.com

This link was posted on Fred Miranda's site...
http://www.outbackphoto.com/artofraw/raw_03/essay.html

Ron Reznick said:

Leon, I shoot NEF only. I converted to TIF in Capture 3.5.

Click to expand...

Okay, the right way.

But I think you are taking the wrong subject for demonstrating the capablities from Capture 3.5. It seems it is a crop of a bird-image or something. So a very fine structured area of "feathers". Isn't it (it is your picture)???

The structure is not seen when shot at 6MP. The downsampling of 4000 pix horizontally shall hide the structure and hide artifacts as well.

Using the full resolution you shall have more artifacts by the more high detail resolution and the problems of a bayer pattern anyway in this special subject (structure of feathers). I can even tell you that if you are using a real more high resolution CCD in a Bayer pattern (e.g. Kodak or Sinar digital backs), you shall see even more artifacts by the rendering of all details from this subject structure. As opposite example, if you shall use a 2.1 Megapixel camera, I guess you shall not see artifacts at all, because the low resolution can not even handle the details, so can not interference with the structure at all.

Do a 2.1 Megapixel CCD have a better quality by the absence of artifacts by this structure than 6 or 10 MP??? No, you are only using the wrong examples.

--
Leon Obers

The 5MP looks better to my eye anyway. There seems to be visible artifacting in the 10MP and the noise in the shadow is more apparent. This was also true when I tried it in Bibble..at least the increase in noise. But it's another way to capture an image and no doubt will have applications for certain users/situations. Better to have the option than not!

Abe said:

Here's what I did...

Each image was opened with Capture 3.5, and then sent to Photoshop
7 -- The first at 5MP and the second at 10MP.

I then resized the 5MP image to the same dimensions as the 10MP
image using Photoshop's resampling tool so I could compare
Capture's 10MP image output to Photoshop's.

No image sharpening or tweaking was done to either image. Both
crops are at 100% magnification.

Here's the resized 5MP image:



Here's the Capture 10MP image:



Capture's 10 MP output is of much better quality in my opinion.

--Abe

http://www.abekleinfeld.com

Click to expand...

--
http://www.profotos.com/pros/profiles/index.cfm?member=612

Leon, there are issues with moire, color aliasing and interpolation artifacts with the D1x image no matter which way you process it. This is one reason why I changed to the D1h.

The 10MP output from Capture exacerbates moire, and the interpolation artifacts on shallow-angled edges are also increased. I have not yet tested problems with primary reds and blues, but I do not expect this issue to be improved, just changed in some manner.

I chose that subject (yes, it is a Black-crowned Night Heron and it is my shot) specifically because it shows multi-angled curved and straight fine detail, and this sort of thing is perfect for illustrating two of the major problems with the D1x image. Keep in mind that I have taken and processed 250,000 shots with D1-series cameras and have a good feel for what they do and how to get the most out of them. When I am testing something with a camera (just as when I am testing something in the engineering field I am involved in), it is likely that I will select subject matter that will allow me to illustrate problem areas rather than avoid them.

I pulled the images earlier, but I am posting this particular one again.



Ron

Leon Obers said:

Ron Reznick said:

Leon, I shoot NEF only. I converted to TIF in Capture 3.5.

Click to expand...

Okay, the right way.
But I think you are taking the wrong subject for demonstrating the
capablities from Capture 3.5. It seems it is a crop of a bird-image
or something. So a very fine structured area of "feathers". Isn't
it (it is your picture)???
The structure is not seen when shot at 6MP. The downsampling of
4000 pix horizontally shall hide the structure and hide artifacts
as well.
Using the full resolution you shall have more artifacts by the more
high detail resolution and the problems of a bayer pattern anyway
in this special subject (structure of feathers). I can even tell
you that if you are using a real more high resolution CCD in a
Bayer pattern (e.g. Kodak or Sinar digital backs), you shall see
even more artifacts by the rendering of all details from this
subject structure. As opposite example, if you shall use a 2.1
Megapixel camera, I guess you shall not see artifacts at all,
because the low resolution can not even handle the details, so can
not interference with the structure at all.
Do a 2.1 Megapixel CCD have a better quality by the absence of
artifacts by this structure than 6 or 10 MP??? No, you are only
using the wrong examples.

--
Leon Obers

Click to expand...

--
Ron Reznick
http://digital-images.net
http://trapagon.com

Ron Reznick wrote:
Thanks for your quick reply.

Member said:

Leon, there are issues with moire, color aliasing and interpolation
artifacts with the D1x image no matter which way you process it.

Click to expand...

I can imagine. E.g. the different resolution in height and wide has their specific characteristics. Sometimes little "squares" are not square at all.

Every digital imager and their capture software do have specific character. None seems to be the same. (I have tested and worked with several types during about ten years). Every imager to have their advantages and disadvantages not written within reviews.

Member said:

This is one reason why I changed to the D1h.

Click to expand...

When you do not need the resolution, I think the "square" pixels and real not interpolated output dimensions of a D1h do give a better image.
[snip]

Member said:

I have not yet tested problems with primary reds and blues, but I
do not expect this issue to be improved, just changed in some
manner.

Click to expand...

Specially the red channel on a D1X for me seems to be less jaggies into another contrasty color within 6MP mode maybe by the different interpolation method for height and wide, than many other camera's using square pixels (specially the Canon 1D do have intense jaggies within the red).

It could be that those jaggies are more "streched" for the D1X images when height is interploated two times.

Member said:

I chose that subject (yes, it is a Black-crowned Night Heron and it
is my shot) specifically because it shows multi-angled curved and
straight fine detail, and this sort of thing is perfect for
illustrating two of the major problems with the D1x image.

Click to expand...

Still I am curious what other high resolution imagers do give at this point. During many testing of several digital imagers and using more high resolution, artifacts are coming in a way as written in my previous message, just by more catching detail and theirfore giving problems if the subject contain a fine structure (e.g. textiles are also giving those problems).
[snip]

Member said:

When I am testing something with a camera (just
as when I am testing something in the engineering field I am
involved in), it is likely that I will select subject matter that
will allow me to illustrate problem areas rather than avoid them.

Click to expand...

Okay, but it could be that you met about similar problems (ofcourse not totally the same) when using those subjects with more high resolution imagers too.

I am aware the interpolation "trick" with the D1X image is only a sort of compromise. Of course you shall not get the same image as a real 10MP imager. It shall contain some errors not shown in real 10MP images. It is only a way to get at least a way to have better images for several occasions than maybe possible when using the default 5.5 MP and upsampling those images.

I do not have NC 3.5 yet (still using NC 2). E.g. I once tested Bibble. But for me I mostly did find the quality of a 10MP Bibble image not satisfied. (It strengthen artifacts too in a very unpleasant way). Maybe by the difference between horizontal and vertical resolution within such a "big" image, I do like the default image about the same quality in height and wide better when they are about the same, even when the real resolution is more low.

Maybe I shall get the same idea when using NC 3.5.

--
Leon Obers