S602's SuperCCD III interpolation test

darkness · 24th October 2002

Many pple are unconvinced by Fuji's claim that the honeycomb design of the SuperCCD III gives better images and can be interpolated upward to double its number of pixels compared to the actual number of CCD sensors. For me, I'm also rather unconvinced. I set up a simple (and not really very scientific) test to test Fuji's claim.

2 images were taken of this scene, a 6mp Normal and a 3mp Fine (to make file size approx the same, I also shot 1 6mp Fine, but didn't seem very different from 6mp Normal):

I used the 3mp image and interpolated upwards using PS's bicubic interpolation. Below are a series of comparison shots of the 2 images from the 3mp -> 6mp interpolated vs. 6mp (all pics are 200% scaled Nearest Neighbour, saved in lossless PNG-24).

Ok, so here's the "challenge". Compare these pics and determine:
1) Which one is better? The left or the right one?
2) Which one is the native 6mp and which one is the interpolated 3mp -> 6mp.

I will reserve my comments for (1) and reveal the answer for (2) at a later time...

Bluestrike · 24th October 2002

I prefer the right pic!

rueyloon · 24th October 2002

left looks better to me

mylau · 24th October 2002

The Pics on the left hand side look better.

The left side are the native 6mp pics.
The right side are the 3mp --> 6mp pics.

Zerstorer · 24th October 2002

Less artifacts seem in the images on the left. However, wonder why the images on the right seem to have slightly higher contrast.

Should have included 6MP fine results to make it more interesting, the benefits of the higher resolution might be defeated by greater compression.

erwinx · 24th October 2002

i think the Fuji's native resolution is 6Mp. So a 3Mp image is actually downsampled by Fuji from 6Mp to 3Mp.

So you're comparing
(i) 6Mp
(ii) 6Mp -> 3Mp -> 6Mp

left looks a tiny bit better

erwinx · 24th October 2002

Quote:

Originally posted by Zerstorer
Less artifacts seem in the images on the left. However, wonder why the images on the right seem to have slightly higher contrast.

Should have included 6MP fine results to make it more interesting, the benefits of the higher resolution might be defeated by greater compression.

thats a good point. jpeg normal may lead to more compression artifacts, such that the comparison is actually a 'normal' vs 'fine' comparison

reflecx · 24th October 2002

Your test does not prove or disprove anything regarding the honeycomb CCD. This is because the 3M image was a in-camera down sampling of a 6M image in the first place. So all you're comparing is one image vs another which has been reduced then enlarged again.

The Fuji S602 has the highest resolution ratings of all 3M cameras (see DPReview and Imaging Resource tests), beating the Sony S75 which has a Carl Zeiss lens and even the D30 (with a L lens)

The Fuji S2 Pro also wins hands down in the 6M DSLR battle, handily thrumping Nikon's own D100. You can actually see visible differences in real world images.

darkness · 25th October 2002

Actually I'm not sure if the 3mp pic is internally downsampled in camera. After all, the number of CCDs is only 3mp, and not 6mp.

When I compared the pics in photoshop, I overlaid them in 2 layers and toggled the top layer on and off so that I can see the difference. The pic on the right has more sharpening halos compared to the left. It also has more jpeg artifiacts. I think you can see the sharpening halo on the right very clearly in the 2nd pic.

So which one is upsampled to 6mp and which one is native 6mp?? Will reveal it tonight...

Keep the comments coming in!! It will probably affect the way I'll shoot in future. Zerstorer, I also shot a 6mp fine, but it has very little perceivable difference from the 6mp normal (to my untrained eyes).

BTW, if you want more crops from the scene from the other places tell me, I'll set it up.

I originally wanted to post animated gifs to show the difference, but gifs only support 256 colours...

Zerstorer · 25th October 2002

Quote:

Originally posted by darkness
When I compared the pics in photoshop, I overlaid them in 2 layers and toggled the top layer on and off so that I can see the difference. The pic on the right has more sharpening halos compared to the left. It also has more jpeg artifiacts. I think you can see the sharpening halo on the right very clearly in the 2nd pic.

Hmm....are you shooting in Normal sharpness setting?

Maybe the comparison should have been made in between 6MPfine/6MPnormal/3MPfine all at soft sharpness to minimize any sharpening artifacts.

Jed · 25th October 2002

Actually, there is an improvement with the Super CCD, I think the quibble has always been that it's not as much as 2x as Fuji claims.

Your comparison is, as others have pointed out, problematic. You really want to show a 6mp Super CCD image, compared with a 6mp normal CCD image. That will best illustrate the improvement in the Super CCD image.

Zerstorer · 25th October 2002

Quote:

Originally posted by darkness

Keep the comments coming in!! It will probably affect the way I'll shoot in future. Zerstorer, I also shot a 6mp fine, but it has very little perceivable difference from the 6mp normal (to my untrained eyes).

Would have liked you to post it up, coz that would minimize the number of variables. In any test, all settings should be controlled and the only thing varied should be the aspect that you wish to test.

Having different JPEG compression settings simply throws the test off balance, even though it may seem insignificant.

After all you are trying to compare the efficacy of the SuperCCD arrangement at 6MP vs 3MP. File sizes shouldn't be a consideration here. Otherwise this test should be renamed 6MP Normal vs 3MP Fine comparison.

mpenza · 25th October 2002

Quote:

Originally posted by darkness
Actually I'm not sure if the 3mp pic is internally downsampled in camera. After all, the number of CCDs is only 3mp, and not 6mp.

it's downsampled. that's how the SuperCCD technology works.

I used to have the 4700 which doesn't have any resolution that's the size of the CCD (2.4MP).

darkness · 25th October 2002

If any of you are still interested in pursuing this test, pls recommend a setup for me. What I can do is:

- take a 6mp fine
- take a 3mp fine
- downsize 6mp -> 3mp and compare
- upsize 3mp -> 6mp and compare

I dunno where in the world I can find a native normal 6mp CCD pic to compare with the S602, but the S602 will almost definitely lose out with half the number of CCDs.

What else would you like to see (comparison wise)? I can set up the test shot tmr night.

Zerstorer · 26th October 2002

Quote:

Originally posted by darkness
If any of you are still interested in pursuing this test, pls recommend a setup for me. What I can do is:

- take a 6mp fine
- take a 3mp fine
- downsize 6mp -> 3mp and compare
- upsize 3mp -> 6mp and compare

Sounds ok to me. May I suggest leave the sharpness at soft to avoid any artifacting at all.

Also, you may want to look into other forms of resampling such as Lanzcos(irfanview/qimagepro) or Genuine Fractals2 if you can get it.

Quote:

I dunno where in the world I can find a native normal 6mp CCD pic to compare with the S602, but the S602 will almost definitely lose out with half the number of CCDs.

Personally I don't think this is necessary....no one here believes that it can equal a true 6MP anyway. The key point should to find out whether shooting in 6MP is worth the trouble over 3MP.

Jed · 26th October 2002

That's the whole point. If you are trying to prove or debunk Fuji's claim that the Super CCD is equivalent to twice the number of conventional pixels, then you should do a conventional 6mp to Super CCD post-interpolated 6mp image. I recommended that test because that was how this thread was originally started.

If on the other hand the test is only a limited one to see the difference between the 3mp and 6mp mode on the Fuji 602, then by all means.

darkness · 26th October 2002

Trouble is that there's aren't many cameras with 6mp CCDs around (apart from the DSLRs), and even then it's difficult to compare since the CCD size won't be the same, the DSLRs obviously win hands down due to their low noise levels.

24th October 2002	#1
darkness Member Join Date: Aug 2002 Posts: 757	S602's SuperCCD III interpolation test Many pple are unconvinced by Fuji's claim that the honeycomb design of the SuperCCD III gives better images and can be interpolated upward to double its number of pixels compared to the actual number of CCD sensors. For me, I'm also rather unconvinced. I set up a simple (and not really very scientific) test to test Fuji's claim. 2 images were taken of this scene, a 6mp Normal and a 3mp Fine (to make file size approx the same, I also shot 1 6mp Fine, but didn't seem very different from 6mp Normal): I used the 3mp image and interpolated upwards using PS's bicubic interpolation. Below are a series of comparison shots of the 2 images from the 3mp -> 6mp interpolated vs. 6mp (all pics are 200% scaled Nearest Neighbour, saved in lossless PNG-24). Ok, so here's the "challenge". Compare these pics and determine: 1) Which one is better? The left or the right one? 2) Which one is the native 6mp and which one is the interpolated 3mp -> 6mp. I will reserve my comments for (1) and reveal the answer for (2) at a later time...

24th October 2002	#3
rueyloon Member Join Date: Jan 2002 Posts: 1,527	left looks better to me __________________ 36frames Wedding Photography - http://www.36frames.com rueyloon - http://www.rueyloon.com

24th October 2002	#2
Bluestrike Moderator Join Date: Jan 2002 Location: There lor~ Posts: 4,843	I prefer the right pic!

24th October 2002	#4
mylau Member Join Date: Jan 2002 Posts: 1,801	The Pics on the left hand side look better. The left side are the native 6mp pics. The right side are the 3mp --> 6mp pics.

24th October 2002	#5
Zerstorer Senior Member Join Date: Jul 2002 Posts: 3,610	Less artifacts seem in the images on the left. However, wonder why the images on the right seem to have slightly higher contrast. Should have included 6MP fine results to make it more interesting, the benefits of the higher resolution might be defeated by greater compression.

24th October 2002	#6
erwinx Senior Member Join Date: Jan 2002 Posts: 2,492	i think the Fuji's native resolution is 6Mp. So a 3Mp image is actually downsampled by Fuji from 6Mp to 3Mp. So you're comparing (i) 6Mp (ii) 6Mp -> 3Mp -> 6Mp left looks a tiny bit better

24th October 2002	#8
reflecx Member Join Date: Feb 2002 Location: Bishan Posts: 762	Your test does not prove or disprove anything regarding the honeycomb CCD. This is because the 3M image was a in-camera down sampling of a 6M image in the first place. So all you're comparing is one image vs another which has been reduced then enlarged again. The Fuji S602 has the highest resolution ratings of all 3M cameras (see DPReview and Imaging Resource tests), beating the Sony S75 which has a Carl Zeiss lens and even the D30 (with a L lens) The Fuji S2 Pro also wins hands down in the 6M DSLR battle, handily thrumping Nikon's own D100. You can actually see visible differences in real world images.

25th October 2002	#9
darkness Member Join Date: Aug 2002 Posts: 757	Actually I'm not sure if the 3mp pic is internally downsampled in camera. After all, the number of CCDs is only 3mp, and not 6mp. When I compared the pics in photoshop, I overlaid them in 2 layers and toggled the top layer on and off so that I can see the difference. The pic on the right has more sharpening halos compared to the left. It also has more jpeg artifiacts. I think you can see the sharpening halo on the right very clearly in the 2nd pic. So which one is upsampled to 6mp and which one is native 6mp?? Will reveal it tonight... Keep the comments coming in!! It will probably affect the way I'll shoot in future. Zerstorer, I also shot a 6mp fine, but it has very little perceivable difference from the 6mp normal (to my untrained eyes). BTW, if you want more crops from the scene from the other places tell me, I'll set it up. I originally wanted to post animated gifs to show the difference, but gifs only support 256 colours...

25th October 2002	#11
Jed Senior Member Join Date: Jan 2002 Location: UK Posts: 3,077	Actually, there is an improvement with the Super CCD, I think the quibble has always been that it's not as much as 2x as Fuji claims. Your comparison is, as others have pointed out, problematic. You really want to show a 6mp Super CCD image, compared with a 6mp normal CCD image. That will best illustrate the improvement in the Super CCD image.

26th October 2002	#16
Jed Senior Member Join Date: Jan 2002 Location: UK Posts: 3,077	That's the whole point. If you are trying to prove or debunk Fuji's claim that the Super CCD is equivalent to twice the number of conventional pixels, then you should do a conventional 6mp to Super CCD post-interpolated 6mp image. I recommended that test because that was how this thread was originally started. If on the other hand the test is only a limited one to see the difference between the 3mp and 6mp mode on the Fuji 602, then by all means.