HowTo? Tell how much resolution lost with DUAL_ISO?

Started by l_d_allan, September 19, 2013, 04:29:45 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

l_d_allan

Quote from: Audionut on September 14, 2013, 03:26:36 AM
The amount of resolution loss is directly related to the distance between ISOs.

DR increase is directly related to the distance between ISOs.

Need more DR, increase separation between ISOs, also suffer increased resolution loss, etc.

Is there a way to tell what the "effective resolution" is when using DUAL_ISO?

With Adobe Bridge, the metadata with my 5d2 continues to show nearly 5616x3744 resolution in the CR2 and the generated DNG (from cr2hdr.exe), regardless of whether I use 100/200, 100/400, 100/800, or 100/1600.

Is that "as expected"?

I also tried with Canon's DPP, and with RawDigger to see if they reported a significantly different resolution. No difference.

BTW, I noticed with RawDigger, that the .CR2 files had max channel values of about 14,740 (same as when not using DUAL_ISO). However, the generated .DNG from 100/1600 had max channel values of about 57,160. I suppose that is the blend of 100/1600? Ingenious / magic. Well done.

a1ex

LOL

Take a 640x480 image, resize it to 5616x3744, and then open it in Adobe Bridge. What does it show?

on-topic: use raw zebras, they show where's fullres and where's halfres.

Doyle4

at 100% crop it can be noticeable if you have a normal pic next to it, but honestly... don't worry about it.

dmilligan

Quote from: Doyle4 on September 19, 2013, 05:25:56 PM
but honestly... don't worry about it.

Disagree. Loss of 50% of vertical resolution in parts of the image is a very important consideration. Esp. in applications where the full resolution of the images is important such as print. Yes if never look at your pictures on anything besides a computer monitor then it won't matter.

"Effective resolution" will vary over the image and from image to image. You will loose resolution in areas of the image that are overexposed in higher ISO and in areas that are underexposed in the lower ISO. If you think about this should make sense. You basically have to throw out overexposed or under exposed pixels, so the result will be than you get half resolution in areas of the image where this happens.

Doyle4

Quote from: dmilligan on September 19, 2013, 06:45:13 PM
Disagree. Loss of 50% of vertical resolution in parts of the image is a very important consideration.

Why would you use Dual Iso if the image will be important? but instead do an actual HDR image with balanced look and not radioactive look... that would make much more sense to me. All my Dual Iso images are resized to half so resolution loss isn't noticeable.

I see what you are saying... but i dont worry about it.

dmilligan

Quote from: Doyle4 on September 19, 2013, 10:39:28 PM
Why would you use Dual Iso if the image will be important? but instead do an actual HDR image with balanced look and not radioactive look... that would make much more sense to me.

I actually would not use dual iso most of the time and would use a bracketed exposure like you say - I think of dual iso as a last resort - but sometimes a bracketed exposure is not possible, esp. when there is a lot of motion in the scene and you would get ghosting in your hdr, or you don't have a tripod handy.

Quote from: Doyle4 on September 19, 2013, 10:39:28 PM
All my Dual Iso images are resized to half so resolution loss isn't noticeable.

What!? If you are resizing to half resolution (vertical and horizontal I assume, to keep the same aspect ratio) then you are throwing away perfectly good resolution in your midtones and you are throwing away horizontal resolution. Resolution loss would actually be MORE noticable by you doing this.

l_d_allan

Quote from: dmilligan on September 19, 2013, 06:45:13 PM
Disagree. Loss of 50% of vertical resolution in parts of the image is a very important consideration.

My "use case" tends to be panos of 180 to 360 degrees FOV. There tends to be plenty of resolution, but dynamic range is often high. I'm trying to use HDR less and less.

  • 14 mm Samyang on full frame ... 4-7 frames in portrait mode results in horizontal resolution of 10k to 14k
  • 8mm Samyang fisheye on full frame ... 3-5 frames in portrait mode results in horizontal resolution of 7k to 10k
  • With an interior involving two rows of 7 frames when using the 14mm, an HDR becomes 42 frames.
  • http://berean.zenfolio.com/uwp

Doyle4

Quote from: dmilligan on September 20, 2013, 01:00:59 PM
I actually would not use dual iso most of the time and would use a bracketed exposure like you say - I think of dual iso as a last resort - but sometimes a bracketed exposure is not possible, esp. when there is a lot of motion in the scene and you would get ghosting in your hdr, or you don't have a tripod handy.

What!? If you are resizing to half resolution (vertical and horizontal I assume, to keep the same aspect ratio) then you are throwing away perfectly good resolution in your midtones and you are throwing away horizontal resolution. Resolution loss would actually be MORE noticable by you doing this.

I see what you mean now, thin were thinking the same on the first part.

Yeah i resize as images are either printed at A4 max or for the web.

Sorry if some reply's came across hostile, wasnt meant to be read that way.

Audionut

Quote from: l_d_allan on September 19, 2013, 04:29:45 PM
BTW, I noticed with RawDigger, that the .CR2 files had max channel values of about 14,740 (same as when not using DUAL_ISO). However, the generated .DNG from 100/1600 had max channel values of about 57,160. I suppose that is the blend of 100/1600? Ingenious / magic. Well done.

The DNGs are 16bit.

Quote from: dmilligan on September 19, 2013, 06:45:13 PM
You basically have to throw out overexposed or under exposed pixels, so the result will be than you get half resolution in areas of the image where this happens.

We're talking 1 frame images here, if you can HDR with multiple images then it's a non-issue.  But where you need HDR from a single frame, then dual ISO cannot be beat, despite any unwanted side effects.  Reduced resolution in highlights/shadows will always be better then no captured detail in those areas  ;)

The trick is working out where to set your gains (ISOs) to minimize the effect of the resolution loss.  Ideally, you set your recovery ISO to only blow true highlight detail.  Half resolution highlight detail is always better then pure white.  Another way of thinking about it, you lose half of your highlight resolution, but you gain detail in the midtones/shadows.  Detail that would otherwise be pushed to far into the shadows (increased noise/reduced post tonal adjustment), or detail that would otherwise be pushed into the noise floor.

Audionut

It comes in handy for fixing exposure errors also.

Here is a shot that's around 2 stops under exposed.


Shot @ 100/400 - No NR


You're going to have a hell of a time dealing with the noise in a shot like this in a standard CR2.