Looking for insight on histograms

[Russ.Carpenter]

I am a long term user of Pixinsight and Astrobin. But these days I'm creating some interesting comparisons between PixInsight and StarTools, anticipating that I'll probably switch to StarTools for post processing.

Many differences between the two program have come to light. In this post, I'd like to focus on histograms. In StarTools, I always use Autodev, in reliance on Ivo's advice. The second screen capture below is a typical histogram from a finished StarTools project: You will notice that the left side of the histogram is almost a straight, vertical line.

The first screen capture is a histogram of the same project processed in PixInsight. Here, the left side of the histogram still rises steeply, but is more of a curve than a straight line. Almost all of my PixInsight projects show a similar finished histogram, and indeed almost all competently processed projects uploaded to Astrobin share the same shape on the left side.

To be clear, I am very happy with the StarTools version. But I am curious to know where the difference in the left side comes from. Thanks for your thinking.

Russ

[admin]

Hi Russ,

It would be useful to be able to see the peaks of the histograms. Also is this a logarithmic representation or a linear one?

From what I'm seeing here, it appears the PI image has a red color cast in the lower dynamic range, while also showing more noise (e.g. dark outliers that extend further from the mean background) than the ST background. The background in the PI image will (probably) appear brighter than the ST image.

Without seeing the peaks of the histograms it is hard to say anything about the shape of the expected "stretched/squished" Poisson distribution (bell curve-ish) around the mode (peak) of the histogram. A stretched image will show the left part of the bell curve squished (narrower) compared to the right part. Pixels below the interstellar background (arguably the mode/peak of the histogram) should ideally be fewer (unless intentionally left there for aesthetic purposes - some measure of noise grain is not always objectionable), as they are lower-than-real background values and therefore are anomalous/outliers. A non-linear stretch will already make them appears less conspicuous. StarTools, however, gives you precise control over them in modules like AutoDev (Shadow Linearity) and Contrast (Dark Anomaly Headroom), allowing you to define how this specific part of the histogram gets "compressed" to make room in the dynamic range for more "interesting" detail (e.g. pixels a the mode/peak or further to the right).

Hope this helps!

[Russ.Carpenter]

StarTools, however, gives you precise control over them in modules like AutoDev (Shadow Linearity) and Contrast (Dark Anomaly Headroom), allowing you to define how this specific part of the histogram gets "compressed" to make room in the dynamic range for more "interesting" detail (e.g. pixels a the mode/peak or further to the right).

Hi Ivo,

As usual, your reply was very intriguing. The quote above describes what I thought might be going on. It takes a while to get used to Autodev, and especially to place the ROI correctly, but the results are remarkable. I need more practice on setting Shadow Linearity and Dark Anomaly Headroom, but already the improvements I see in details and midtones are eye-catching.

Russ

[happy-kat]

this post illustrates pulling out faint stuff
viewtopic.php?f=11&t=1743&p=7490&hilit=autodev+orion