Example of Pix Insight HDR Wavelet Process

[roundycat]

TJ, thanks for the M42 re-work with high pass. Not knowing your original picture's resolution I guessed at the radius and got it wrong, maybe 20 would have been better.

The point of the exercise was quite simple. Somewhere along the line there is a limit to the number of ways you can write a sharpening kernel. We are, after all, merely changing the relative brightness of adjacent pixels. That's it. I felt that the M42 pic was pointlessly overdone and wondered how it would fare being overdone to a similar degree in PS.

There is a degree of proof here that you can muck up a picture equally well in PI or PS. It would be nice to see the two compared on a level field where the object was the best possible result in both cases.

What I'm really wondering is how easy is it to replicate wavelet processing in PS.

Dennis

[Tim]

Oh I see.

I should probably have pointed out that my M42 at least was HA+O3+S2 +RGB as I recall, and not a straight lrgb.

A good test of PS v PI would be on an object like M31 or M42 as regards HDR wavelets. I can supply unprocessed data if anyone fancies it? The aim being to ensure every ounce of available detail is displayed, and not lost somewhere along the lines, if that counts as the best possible result?

[MartinB]

A good test of PS v PI would be on an object like M31 or M42 as regards HDR wavelets. I can supply unprocessed data if anyone fancies it? The aim being to ensure every ounce of available detail is displayed, and not lost somewhere along the lines, if that counts as the best possible result?

I'm up for this but I might have to produce 2 images - one with every ounce of detail immediately apparent and something a bit softer with the finer details a bit further back in the mix.

[roundycat]

from the point of a strict comparison I am not convinced that the type of data, ie, rgb, lrgb, narrowband etc is relevant. Careful processing brings out the fine detail regardless of where it is.

For me a good test would be to compare part of a decent nebula such as M42 or the Rosette using wavelets, highpass and plain old unsharp masking (not USM sharpening filter).

Perhaps a 16 bit TIFF could be made available cropped down to about 600 pix square to avoid massive downloads.

Dennis

[Tim]

ooooh ack 16 bit

[roundycat]

my mention of 16 bit was for two reasons. First, it makes for a smaller file size and Second, it doesn't lend any mystique to 32, 64 or 500 bit image depth - we are all limited to the display medium of about 7 bit.

Furthermore, if you think about your camera it is unlikely to exceed 16 bit, none do that I know of. So, the most meaningful comparison would be to take a well exposed frame - hence the need for a bright target - and process it using 16, 32 or 64 bit as you choose. Just remember that any bit depth greater than 16 is likely to be artificial.

Once the various pictures are processed they can all be saved as 8 bit jpegs for display and only then can we make a really meaningful judgement as to which is the best.

To allow any kind of subjective assessment to play with the results is to ensure the results are meaningless.

Dennis