Example: ISO 800 lowlight+PSCS+NeatImage (10D)

[who10]

I’ve noticed several posts lately raising points regarding low light experience with ISO 800 or higher, the relative benefit of Adobe Camera Raw vs other converters, whether to apply noise reduction before or after sharpening, etc.

Here is an example of one work flow in PSCS (if anyone finds this interesting I’ll post a second comparing C1 Pro). I made exactly four corrections in total. The shot was taken in a dimly lit auditorium under incandescent floods as follows:

10D
ISO 800
16-35Lf2.8 @ f2.8 (five meters from subject center)

The first frame in the image top left is the Raw file converted with no adjustments. The second is the result of correcting the color temperature, the top right after increasing exposure +1.10 => no other raw file processing. This third image was saved as a 16bit tiff file and loaded into Neat Image (using a custom ISO 800 1/60th f2.8 EOS-10D noise profile). The last step was to sharpen the final montage for web-display using Photoshop.

Hope it’s strikes some as interesting… David

http://www.haskellct.com/photos/ExampleWtext.jpg

[CyberDyneSystems]

Wow,. the Neat image did a fine job!

[Belmondo]

I'm impressed. I'm obviously going to have to go back a work some more with Neat Image because I never got results like that. Maybe I should have printed the manual and read it. :oops:

[robertwgross]

I've found Neat Image to be very effective as long as a few conditions exist.

(1) I need to manually set it up. No automatic batch processing.

(2) I need to find a piece of fairly normal background without detail, and it needs to display the normal noise or color noise or grain. That background can't be abnormally light or dark.

For every ten images that I process with Neat Image, probably two will result in no improvement. Probably two will result in very nice improvement. The rest are OK.

---Bob Gross---

[who10]

I never got results like that. Maybe I should have printed the manual and read it. :oops:
:shock: Stop... there will be no reading of the manual ! :wink:

In one of my weaker moments (after general frustration following above advice)... I did stumble across a reference to making "custom profiles" for one's camera -- too much work said I, then - ahhHa, manual says one can simply load the test target onto one's monitor, set the focus to infinity, fill the frame with the test pattern and shoot the screen image - that sounded pretty easy...

Viola, 10D noise finger print for said aperature/ISO and shutter setting. You then run one analysis, save the profile and load it for future use. My experiments seem to show ISO and aperature are the most important attributes, shutter speed rather secondary. If you follow this recipe it matters less when you remove the noise - but I generally remove noise immediately after raw conversion just before other post processing (but always BEFORE sharpening).

The process change really improved the consistency I was getting from NI. Prior to this I did verbatim what Bob Gross describes - with similar results. Not every image is as dramatic, but my "hit rate" is much better and I can run batch again (all images with same ISO/aperature seems to be close enough). I've done as many as 50 in a batch with excellent results.

Give it a try... I'd be interested if others find the same results.

David

[Scottes]

I've always been pretty happy with Neat Image. But that custom profile thing seems like a winner. It might be a little bit of a pain, but seems like it would be worth it.

[maderito]

I've found Neat Image to be very effective as long as a few conditions exist.

(1) I need to manually set it up. No automatic batch processing.

(2) I need to find a piece of fairly normal background without detail, and it needs to display the normal noise or color noise or grain. That background can't be abnormally light or dark.

For every ten images that I process with Neat Image, probably two will result in no improvement. Probably two will result in very nice improvement. The rest are OK.

---Bob Gross---
Excellent points. Shots that really need serious noise removal often have very uneven lighting conditions - and thus uneven noise characteristics. Although the NI noise removing algorithms are quite sophisticated, they can't compensate for all the factors that can vary in one shot (color temperature, exposure, color shifts, etc.).

My results are better than Bob's 2/10, but I shoot probably shoot different types of subjects/scenes which have more areas to sample for noise characterization.

Batch processing is a real no-no for critical work under demanding shooting conditions.

[Scottes]

I've found that I can get good to very good results if I can set the "analyze" box over an area containing both light and dark areas. It bothers me when I see the sliders in a straight line as I know that noise won't be even from dark to light.

If necessary, use the history brush to remove the noise removal from detail sections, or use some creative sharpening to get back noise-less detail.

Also be sure to adjust final sharpening when dealing with noisy images. Setting a higher threshold will reduce the sharpening of noise. Of course you have to adjust the amount accordingly, and it always depends on the image.

This image was ISO 800:
http://www.itsanadventure.com/postimages/BlueJay_7371-2.jpg

And a 100% crop, NR & processing but no sharpening:
http://www.itsanadventure.com/postimages/BlueJay_7371-1-crop.jpg

Now sharpening. Original is on the left, the middle shows USM at Amount 150%, Radius 1.1, Threshold 0, and finally the right side shows USM at 225%, 1.1, Threshold 6. The threshold was increased to reduce the sharpening of noise, but then the amount has to be increased to bring it back to approximately the same sharpness.
http://www.itsanadventure.com/postimages/BlueJay_7371-nr.jpg

This shows at 100x100 crop of the lower left corner of the 2 sharpened images, at 300% zoom. 150/1.1/0 on the left, 225/1.1/6 on the right.
http://www.itsanadventure.com/postimages/BlueJay_7371-noise.jpg

As you can see the USM Threshold setting has a dramatic change in how the noise appears in the final image. Yes, it changes the image a bit - the lower threshold does have slightly more contrast - some of the smallest feather & "whisker" details are lost with a threshold of 6. But the noise is dramatically reduced.

Ideally speaking I would sharpen this image selectively - easy to do with such a background - through the use of a mask or by de-sharpening the background with the history brush.

[gmitchel]

Threshold for USM will affect the appearance of noise.

I suggest you work on the USM settings for the bird. Both show some very obvious sharpening artifacts, probably compounded by the JPEG format.

The post with the kids used an Exposure slider increase of more than a stop. You need to be careful with the Exposure slider in ACR II. If you increase EV by 2/3-3/4 stop or more, you will almost certainly see increased noise in the shadows and probably posterization. Of course, the more extreme the Exposure slider setting, the more obvious the artifacts.

In ACR II, you are working with an image that has not been gamma-adjusted. It is still linear with regard to brightness values. As you increase the Exposure slider, the shadow information spreads out, which makes noise more apparent.

Cheers,

Mitch

[Scottes]

I suggest you work on the USM settings for the bird. Both show some very obvious sharpening artifacts, probably compounded by the JPEG format.

Yes, I know. It was over-sharpened a bit to exaggerate the noise.