ISO and downsampling

[Tedphoto]

I have a question: It used to be that lower mp cameras had better high iso performance because of larger photosites on the sensor. Now I am reading that is no longer true because a down sampled image from a larger mp camera can take advantage of the increased photosites to create an image with less noise when downsampled. I’m wondering if this only applies to an image that is processed as a jpg in the camera? Is there a way this would also work when editing a raw image in LR or Canon’s software?

[johnrmoyer]

Since the noise from high ISO seems to mostly be extreme values when compared to surrounding pixels,  a median filter will make the noise more like the surrounding pixels which makes the noise less visible, but also blurs the image. Down sample and sharpen after a median filter will result in a lower resolution image that sometimes will look better than the higher resolution original. If a higer resolution image is needed, then this will not be useful, but if a lower resolution image can be used then it is worth trying. Gimp or graphicsmagick will do a median filter and resample.

Here is a down sampled noisy 25600 
ISO image: https://www.rsok.com/~jrm/2021May26_birds_and_cats/2021may22_fox_IMG_1312c.html 

I used graphicsmagick on a Debian Linux computer to do the downsample after converting the raw CR3 file to JPEG using DPP on an iMac. DPP did a lot of noise reduction leaving a very soft image but did well at preserving colors. The downsampling took an unusuable fuzzy and noisy image and changed it into an interesting but not perfect images of a fox eating bird seed, grape jelly, and a strawberry. Illumination was a porch light and the photo was made from inside my house though a double pane window which further distorted and softened the image. Down sampling also made some of the motion blur from the slow shutter speed and the fox moving less obvious.

The graphicsmagick command line used was (extra quotes around ':' to keep it from turning into emoticon): 

gm convert -verbose IMG_1313c.JPG -resize "37.5%" -unsharp 0x1 -mattecolor black -frame 16x16 -font helvetica -fill white -draw "text 4,1228 'Gray Fox (Urocyon cinereoargenteus), May 22, 2021, Copyright http://www.rsok.com/~jrm/ All rights reserved.'" -define 'jpeg:dct-method=float,jpeg':'optimize-coding=true' -interlace line  -quality 97 2021may22_fox_IMG_1313c.jpg

F Number	5.0
ISO	25600
Shutter Speed Value	1/83
Focal Length	286.0 mm
Focus Distance Upper	10.47 m
Focus Distance Lower	8.99 m
Lens Model	EF100-400mm f/4.5-5.6L IS II USM

[ebiggs1]

Kinda looks like an over use of the clarity slider. 

[johnrmoyer]

Clarity slider was at 2.
Unsharp mask was 2.0,4.0,4.0 with additional unsharp mask applied after down sampling.

Digital Lens optimization was at 40.8.

Noise reduction was at 12.0 for both Luminance and Chrominance.

Setting Clarity slider to 0 instead of 2 makes very little difference. The image was very soft because of 25600 ISO, noise reduction, motion by the fox, slow shutter speed, and dual pane window glass with reflections. The point was to show that downsampling can be useful for rescuing a photo made under difficult conditions.

The Fox moved less and posed better in this one: 

---

@ebiggs1 wrote:

Kinda looks like an over use of the clarity slider. 

---

[ebiggs1]

"Clarity slider was at 2."

I understand. I figured it was set fairly low. I merely remarked that I think you could get the same appearing photo with the clarity slide set pretty high. In photography there is no free lunch. You always give to get.

Many, many ways to achieve the end result.

[Waddizzle]

The graphicsmagick command line used was (extra quotes around ':' to keep it from turning into emoticon): 

gm convert -verbose IMG_1313c.JPG -resize "37.5%" -unsharp 0x1 -mattecolor black -frame 16x16 -font helvetica -fill white -draw "text 4,1228 'Gray Fox (Urocyon cinereoargenteus), May 22, 2021, Copyright http://www.rsok.com/~jrm/ All rights reserved.'" -define 'jpeg:dct-method=float,jpeg':'optimize-coding=true' -interlace line  -quality 97 2021may22_fox_IMG_1313c.jpg

F Number	5.0
ISO	25600
Shutter Speed Value	1/83
Focal Length	286.0 mm
Focus Distance Upper	10.47 m
Focus Distance Lower	8.99 m
Lens Model	EF100-400mm f/4.5-5.6L IS II USM

-----------------------------------------------------------

I never new that 1/83 shutter speeds were possible with a Canon camera body.  Those settings are recipe for a poor image.

[ebiggs1]

I tought them odd too. But it did produce a nice photo. However, I believe I could do something similar with the clarity slider in LR.

[johnrmoyer]

It was aperture priority and auto shutter speed and I had guessed that 25600 was the highest ISO I could get by with. So far as I know, with aperture priority the shutter speed might be set automatically to any arbitrary number. The fox came while we were watching television and only stayed a couple minutes. The porch light was on, but else it was completely dark at the tree stump where I feed birds. I went into the other room to get my camera and was able to return without scaring the fox. The photo was made through double pane window glass which provided distortion and reflections. The point of posting was to demonstrate that down sampling can rescue a photo made under difficult conditions. The original question was about downsampling and high ISO noise and not about making a photo under optimal conditions so far as I can remember.

Maybe if I had paid my foxy model better I could have gotten a studio shot with proper lighting 🙂 The only pay was bird seed and fruit on a tree stump.

"

Measured EV                     : 3.12

Measured EV 2                   : -16.5

" from Exif data.

---

@Waddizzle wrote:

-----------------------------------------------------------

I never new that 1/83 shutter speeds were possible with a Canon camera body.  Those settings are recipe for a poor image.

---

[ebiggs1]

"... bird seed and fruit on a tree stump."

Perhaps I should have said there is no free lunch, ... most of the time! 

[ebiggs1]

"I used graphicsmagick on a Debian Linux computer to do the downsample after converting the raw CR3 file to JPEG using DPP ..."

You may as well be talking Chinese. I use and have never really used anything else besides Photoshop and Lightroom. I do have DPP4 installed on my Windows 10 computer but rarely use it.

[johnrmoyer]

I think that the "Digital Lens Optimization" in DPP is worth it if one is using both a Canon lens and body. Else, I would likely use the free rawtherapee for raw development. I frequently use gimp and graphicsmagick and exiftool. I no longer use any Adobe or Microsoft software. I had heard that either lightroom or photoshop is able to do something similar to a Richardson/Lucy deconvolution to remove diffraction blur. I like the way DPP does this, especially for small aperture images. Since I have retired, I almost entirely am able to avoid using Windows. I have written PC BIOS, embedded firmware, RADAR algorithms, and modeling of chemical accidents. I know about sensor data and algorithms. All sorts of sensors can produce noisy data.

Now that I am retired, I am trying to learn photography.

---

@ebiggs1 wrote:

"I used graphicsmagick on a Debian Linux computer to do the downsample after converting the raw CR3 file to JPEG using DPP ..."

You may as well be talking Chinese. I use and have never really used anything else besides Photoshop and Lightroom. I do have DPP4 installed on my Windows 10 computer but rarely use it.

---