blackest Loc: Ireland

Ok this is a pretty common scenario light is limited.

Ideally i wanted f7.1 1/250th and ISO 100 but I had to use f7.1 1/250th and ISO400 Why?
I needed enough dof f7.1
I needed to avoid image shake so i used 1/250 (focal length was 210mm)
I was forced to use iso 400 since the light level was 11.7 EV as my subject was in shade.

If i had 13.7 EV then i could have used base iso
raising iso reduces dynamic range and increases noise. Maybe i could have shot 2 stops under exposed and restored the exposure in post processing...
To be fair loss of dynamic range wasn't an issue the highlights were not that bright and nothing was blown.

By raising ISO though I was boosting signal and noise.
for an 8 x 12 print at 300DPI i need 2400 x 3600 pixels my sensor has more than double that.
noise is random and each pixel is going to vary in how much noise they carry.
So what if I downsize by 50% effectively making my pixels 4x the area receiving 4x the photons the noise across the 4 pixels will be averaged out, there still will be some noise of course but lower than across the 4 separate pixels e.g if the noise was 3 5 7 4 across the 4 pixels averaged it would be 4.75 and not so noticeable.
Of course that will reduce image resolution but scaled down it's still over 300PPI so the detail lost is invisible, without the scaling i wasn't going to see it anyway.

So why not supersize my pixels and improve the noise levels?
maybe applying a slight crop to 7200x4800 or (3600 x 2400 after downsizing)before printing might help?
There are a few methods for down sampling, bicubic sharper is probably better than just using nearest neighbor.
Is this a valid method to improve image quality ?

What do you think?
Maybe its better to apply noise reduction to the darker pixels only where the signal is weakest, in brighter areas you may be just blurring detail.

How to obtain best image quality, that you can see?

martinfisherphoto Loc: Lake Placid Florida

Silly Question, I know most of us are anal and our worse critics, but how are you seeing Terrible Quality at ISO 400. I know your question is about how to obtain the best possible result, but enlighten me. If viewed at a normal size on computer and not blown up to 400% how are you seeing ISO400. I recently went to Costa Rica and came back with some of my best captured material so far.. Problem being, under poor lighting conditions, deep jungle or misty cloud forest, I have some grain in my photos. Some times we are stuck with what we have. If I down size to show off on the web, and reduce noise that doesn't help when I want a 24 x 36 printed image, it is what it is.. I am my worst critic, but I am starting to understand, it's all about the image. If the right moment is captured, noise won't stop you from Feeling It.....Does your image have to be Perfect for you to Feel It????

Gene51 Loc: Yonkers, NY, now in LSD (LowerSlowerDelaware)

You are thinking about this too much, John.

I recently took this using a D810 at ISO 4000, 1/800 and F7.1 600mm. I downsampled to 2048x1973 from 36 mp. Applied a little NR, masking, sharpening, clarity etc - to get it looking like the first image - the second image I fiddled a bit with microcontrast, sharpening and masking, and noise reduction. The difference is subtle on screen, though much more noticeable when printed.

If you are using a camera that is ISO invariant, you can raise ISO without really adding more noise, or use base ISO and underexpose up to 5 stops and end up with the same result.

(Download)


(Download)

RichardTaylor Loc: Sydney, Australia

I shoot a lot of high ISO images whilst covering indoor classical music concerts.
Usually ISO 3200 to 12800 on a M4/3 camera..
Images are usually used for web publication or smaller prints (up to 10x8 or A4).

Downsizing to 1920x1080 hides a lot of problems.
I also use Topaz software to reduce the noise (especially for the performers who are almost always wearing black). Also the slight softening, when doing the noise reduction, tends to make the senior ladies faces a little softer (however I may sharpen the eyes).
In extreme cases I will do a B&W convesion (especial for the male portraits).

Pablo8 Loc: Nottingham UK.

Couldn't you have used a tripod/ monopod to steady the camera, therefore using the 'slower' shutter speed as the light level dictated?

blackest Loc: Ireland

Photographing a rock maybe, but even plants will move in the slightest of breezes.

http://mecastronics.free.fr/ExposureLightValueVCam.html

This is ugly but lets you figure EV, shutterspeed, iso and f-stop.

tonyantony

photography should be fun...just point and shoot

f8lee Loc: New Mexico

Seems to me there's more complexity here - since the IR/heat photos that create what we call noise are being sensed by photo sites, each with its own color filter above, and the demosicing process combines all that via voodoo (since there are no specific lookup tables as there are with, say, JPEG RGB files), is it not questionable as to how effective all this folderol will be? To use your example of adjacent pixels (which really would have to be the photo sites on the imaging chip, again, each with a single color filter above) then what happens if the "7" value falls on a Red filtered site, versus a Green filtered site?

And no, I don't have an answer...my point is that this may not be as straightforward an issue as you might think.

CamB Loc: Juneau, Alaska

When you tried this, what did you discover? I think you are way over thinking this and can get the results you want with a much simpler work flow. Also, how are you getting the noise you talk about at iso400. Even on my crop camera I don't see any difference between iso100 and about iso1000 on 17x25 inch prints. (my favorite size)
Check your camera. Seems like something is out of whack.
...Cam

carl hervol Loc: jacksonville florida

You need to just take picture and quite thinking.

R.G. Loc: Scotland

I may be wrong, but won't the downsizing you're talking about happen at the printing stage, if it's needed, and be determined by the DPI?

I can see an advantage in doing any denoise at the higher resolution, because then the noise will be smaller, so it'll be easier to treat it with less loss of the micro-detail that you may want to keep.

aellman Loc: Boston MA

I'm getting motion sickness.

Gene51 Loc: Yonkers, NY, now in LSD (LowerSlowerDelaware)

You can downsample at any time to see the benefit. I routinely do not resize for printing. If I print in house, Qimage handles resizing and image optimization by default. If I send images to a lab, their RIP hardware/software does the same. DPI specifically relates to printing resolution. Image resolution, and downsizing from 36 mp to 4 mp can (and does) happen independently of printer resolution.

blackest Loc: Ireland

The resizing does lose some resolution, noise is high frequency most of what we see in the world is not. My camera effectively is at 600dpi for an 8 by 12 print. I can't see 600dpi , i can't see 300dpi either so by resampling down to 300dpi i have lost some image detail but i was never going to see it anyway.

One method for improving noise is stack a series of images and the variation between the stacked pixels is pretty much just noise.
Richard earlier in the thread has used resizing to pull down noise levels.

https://blog.kasson.com/the-last-word/noise-effects-of-photoshop-bicubic-sharper-downsizing/

The worse the lighting the larger the effect of noise and we know larger pixels handle low light better than smaller pixels by the resize we effectively make our pixels 4x bigger.

R.G. Loc: Scotland

It seems to me the moral of the story is that you don't want to use a higher resolution than is necessary for the specific end result that you have in mind.

Your link refers to bicubic reduction. What are the alternatives and does bicubic provide the best noise reduction? On1 uses fractal-based upsizing. Is there a downsizing equivalent?