Pages: [1]   Go Down

Author Topic: How to overcome D3 low-noise envy  (Read 3631 times)

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
How to overcome D3 low-noise envy
« on: December 01, 2007, 11:01:16 am »

I believe the Nikon D3 is already shipping but I can't find any clear and unambiguous comparisons between it and the Canon 5D at high ISO..

Although the 5D produces images at ISO 3200 which are pretty impressive regarding low noise, there is inevitably a noticeable loss of resolution, a slight softening of the image, as well as pretty obvious noise in the lower mid-tones and shadows if there's the slightest bit of underexposure.

When the D3 was first announced and much was made of the fact it boasted an ISO 6400, I assumed it would be a 'false' ISO like the 5D's ISO 3200. I was surprised to learn later that it is in fact a regular ISO settings and the false ones (post A/D amplification) are ISO 12500 and 25000.

If ISO 12500 and 25000 are just ISO 6400 underexposed by one and two stops respectively, then any noise comparisons should be between the D3 at ISO 6400 and the 5D at ISO 1600, 3200 and 3200 underexposed one stop.

The comparison images should be carefully examined for noise and resolution. There's not much point in having lower noise from the camera if resolution suffers.

Okay! So much for the preamble. I fully expect the D3 at ISO 6400 to exceed the performance of the 5D at ISO 3200 underexposed one stop (or underexposed 2/3rds of a stop, whatever, depending on ISO accuracy. The 5D's ISO 3200 is actually ISO 4000, so that should be borne in mind during any comparisons.)

So, how to get your 5D performing at least as well at high ISO as the D3? There is a way, but it only works with motionless subjects. Buy yourself an upgrade to CS3 Extended. It's much cheaper than a D3.

Below is a number of 200% crops taken from the flattened result of 6 stacked handheld 15mm shots (no IS) at 1/20 sec, f16 and ISO 3200, compared with crops from a single shot from the series.

I'm pleased with the results. Not only is noise dramatically reduced in the stacked image but resolution has also noticeably improved as a result of the lower noise. The inevitable misalignment amongst the images seems to have been perfectly corrected with CS3E's auto-alignment feature.

Okay! No need to mention the obvious. Anyone who owns both a D3 and CS3E will continue to have a low noise advantage.  

[attachment=4089:attachment]  [attachment=4090:attachment]  [attachment=4091:attachment]  [attachment=4092:attachment]  [attachment=4093:attachment]
Logged

sojournerphoto

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 473
How to overcome D3 low-noise envy
« Reply #1 on: December 01, 2007, 04:44:39 pm »

Quote
I believe the Nikon D3 is already shipping but I can't find any clear and unambiguous comparisons between it and the Canon 5D at high ISO..

Although the 5D produces images at ISO 3200 which are pretty impressive regarding low noise, there is inevitably a noticeable loss of resolution, a slight softening of the image, as well as pretty obvious noise in the lower mid-tones and shadows if there's the slightest bit of underexposure.

When the D3 was first announced and much was made of the fact it boasted an ISO 6400, I assumed it would be a 'false' ISO like the 5D's ISO 3200. I was surprised to learn later that it is in fact a regular ISO settings and the false ones (post A/D amplification) are ISO 12500 and 25000.

If ISO 12500 and 25000 are just ISO 6400 underexposed by one and two stops respectively, then any noise comparisons should be between the D3 at ISO 6400 and the 5D at ISO 1600, 3200 and 3200 underexposed one stop.

The comparison images should be carefully examined for noise and resolution. There's not much point in having lower noise from the camera if resolution suffers.

Okay! So much for the preamble. I fully expect the D3 at ISO 6400 to exceed the performance of the 5D at ISO 3200 underexposed one stop (or underexposed 2/3rds of a stop, whatever, depending on ISO accuracy. The 5D's ISO 3200 is actually ISO 4000, so that should be borne in mind during any comparisons.)

So, how to get your 5D performing at least as well at high ISO as the D3? There is a way, but it only works with motionless subjects. Buy yourself an upgrade to CS3 Extended. It's much cheaper than a D3.

Below is a number of 200% crops taken from the flattened result of 6 stacked handheld 15mm shots (no IS) at 1/20 sec, f16 and ISO 3200, compared with crops from a single shot from the series.

I'm pleased with the results. Not only is noise dramatically reduced in the stacked image but resolution has also noticeably improved as a result of the lower noise. The inevitable misalignment amongst the images seems to have been perfectly corrected with CS3E's auto-alignment feature.

Okay! No need to mention the obvious. Anyone who owns both a D3 and CS3E will continue to have a low noise advantage.   

[attachment=4089:attachment]  [attachment=4090:attachment]  [attachment=4091:attachment]  [attachment=4092:attachment]  [attachment=4093:attachment]
[a href=\"index.php?act=findpost&pid=157474\"][{POST_SNAPBACK}][/a]

You can also do that for free with CombineZM - only on 8 bit images, though at 3200 iso that may be fine:)
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
How to overcome D3 low-noise envy
« Reply #2 on: December 01, 2007, 11:53:21 pm »

Quote
You can also do that for free with CombineZM - only on 8 bit images, though at 3200 iso that may be fine:)
[a href=\"index.php?act=findpost&pid=157540\"][{POST_SNAPBACK}][/a]

Maybe you can but I can't. Check out the comparisons below.

The interesting thing for me regarding the new auto-alignment feature of CS3 and CS3E is that it really works. For someone who doesn't like walking around with a tripod, this is a great boon. If you are carrying a tripod and the subject is stationary, then getting low noise images in poor light is no problem. You just use a long exposure at a low ISO.

In this example I've demonstrated where a 1/20th sec exposure was required at ISO 3200, it would not have been possible to hold the camera sufficiently steady at 1/5th or 1/10th second exposures for a sharp result. Stacking to reduce noise was the best option.

CombineZM might be able to deliver some flawless results under certain conditions and with sufficient expertise from the user. However, with these 7 images of my hotel room (I wrote 6 before but it's actually 7) the 2 pass alignment feature of CombineZM provides some improvement over no alignment at all, but it's not a patch on CS3E's auto-alignment which actually appears to create a slightly sharper image than a single noisy one. It's probably not sharper. The detail is just less obscured due to a reduction in noise.

In the comparison below, from left to right, we have CombinZM with no alignment, CombineZM with 2 pass alignment and CS3E with auto-alignment.

The same converted 8 bit TIFs were used with both programs. Stacking 7x16bit images in CS3E requires more memory than I have in my laptop.

[attachment=4103:attachment]
Logged
Pages: [1]   Go Up