Using Neat Image noise removal software - advice needed

Ok so I've got this software package and I've gotten some impressive results with it, but thus far only from my very light tinkering with sliders and not really with any understanding of what I am doing nor the direct effect of what I'm changing is having save what I can get from the preview box.
So I'm out to learn some more about using this software to a better standard than I currently am. For reference my current workflow is outlined below:

1) Open image in Photoshop Elements 6 - process from RAW and then open up in the editor itself.

2) Open the Neat Image plugin interface

3) Use the Device Noise Profile for the 400D for the selected ISO value (profile downloaded of the Neat Image website). Working in colourspace YCrCb (I have no idea what this means)

4) Select an area of the photo for the Rough Noise Analyser - making sure that no warning factors are displayed and that the bar itself has as little/no coloured flecks in it (ie the selection window is clean)

5) Use Auto Fine tune and Auto complete

6) Move to the Noise Filter Settings tab

7) Use Filter and Sharpen Image *

8) Review the preview image and adjust the sharpening sliders experimentally to try and get a pleasing result - again this is another area where I'm operating blind as to what I am doing; I can only see the result.

9) Click apply and the editing is done.

As you can see there are a lot of stages where I am either totally reliant on the auto settings in the software or simply reliant on a level of blind luck as I adjust the settings to get a decent appearance. If anyone has any pointers I'd greatly like to have your input in this matter.


* I am aware that I should strictly speaking use filtering only on the first pass and then run the software again as a final process for sharpening only (and use layermasks in the main area of elements to mask the effect of sharpening off from areas that do not need it).

Steps 3,4 and 5 are redundent. The downloaded profile was made by somebody doing 4 and 5 on a target image and should theoretically be sufficient. However, the converter engine used for the RAW conversion (DPP as opposed to ACR, for instance) and, of course, any editing done in the RAW converter, such as sharpening, contrast, saturation/vibrance, clarity and especially lightening the image, will affect the amount of visible noise in the rendered RGB image. So unless you are duplicating the profile maker's workflow the accuracy of the profile will be less than optimum. On the other hand, for step 4 you need an area of solid color in the image, which may not always be the case. I'd say that if you do have a solid area, 4 and 5 are preferable, but without the sampling area 3 can be used to good affect, especially if in step 6 you are tweaking the settings instead of simply going with the default.

YCrCb is a different way of organizing the image data (as opposed to the more conventional RGB) which separates it into luminance data (Y) and chrominance data (Cr and Cb). This allows Neat Image to apply the noise reduction only to the luminance channel and leave the image colors unaffected.

Regarding sharpening: NR and sharpening are mutually opposed. NR blurs detail together with noise and sharpening does the opposite. Using them together you have to find the best balance, giving just enough sharpening to rein in the effects of the NR. Normally, the accepted advice is to do "capture" sharpening in the RAW converter, enough to cancel the effects of the blurring AA filter in the camera and to give the image a reasonable but not final level of sharpness, with the assumption that the final sharpening (output sharpening) will come at the end of the workflow. I'd advise not sharpening in ACR and regarding the NI sharpening as your capture sharpening.

Thanks for the guidance on things - its certainly brought the program a little more into the light for me.

I've noticed that when I do stages 4 and 5 the bar sliders can often change by a noticeable amount on the controller and that is with a clear sample area. I suspect this is going back to your point about emulating the original users RAW processing method and the fact that I'm not repeating their direct process in the least.

The point about YCrCb moving the noise reduction to the luminescence and leaving colour alone I half understand in that I've encountered such noise reduction methods in the past and seen the good effect they can have on the noise without harming the colour of the shot, so its good to know that the process is already taking place within Neat Image itself.

Moving back to the sharpening point at present I have left the sharpening setting in ACR to its default value and never touched it (at first I didn't even use the tab to realise that it was already making a big difference in removing the softness from the AA filter). With what you say about moving the sharpening totally away from ACR and into Neat Image I take it that you mean I should process the whole shot first - then perform the capture sharpening and then finally the output sharpening. However I worry that sometimes one needs to have a capture sharpened shot to do fine adjustments in the editing stage - so should I move the capture sharpening forward (as I've been mostly unknowingly do in the past) in the process so that I at least have a sharp shot to work from during the editing phase? The other concern is how to fully draw the line with a new tool between capture sharpening and output sharpening - ie not taking the capture sharpening too far in Neat Image and ending up with it already at output sharpening levels early in the editing process?