RGB/K Cursor Readout Values?

[ftherrmann]

I understand that the cursor readout values are virtual within PI.

By example:

I have loaded an image of M17 (32bit fits) with over-exposed areas (ie. saturated)

I've set the readout mode to RGB/K and 8 bits with the intention of being able to read bitmapped equivalents.  (ie. Saturated areas should have a k=~255)

Here's what I do not understand:  When placing the cursor over the saturated areas I'm seeing max values of ~17.  Why is this not showing something closer to the expected ~255?

Thank you,

Fred

[Carlos Milovic]

Are you using an automatic stretch with the STF?

[ftherrmann]

Yes. 

[ftherrmann]

I've tried both STF stretched and un-stretched and with single subs as well as integrated subs. As expected the STF doesn't have an effect on the readout value.  If a STF stretch were to change  the K value wouldn't that be against the "philosophy" of PI?  ie. STF's never change the raw data?

The saturated area (ie. Stars) in CCDStack give the correct bit mapped values (~255).  These same stars in the 8 bit readout mode of the 32 bit fits integrated stack in PI gives me a value in the 40's which is obviously NOT correct.

[Juan Conejero]

What readout numbers do you get for those pixels in the [0,1] range with 4 decimals for example (i.e. if you select the normalized real range)?

[Carlos Milovic]

Of course, STF would not affect readouts. I was just trying to figure out if what you called "saturated" signal was what you perceived as white after the STF.

If it is raw data, there could be 2 reasons why saturated signal is not 1.0:

- The analog to digital converted used less than 16bits to sample the data, and it was stored in 16bits format using the least significative bits. For example, my dslr images have saturated values at nearly 16000, since it is a 14bits ADC.
- Raw data stored in 32bits format may not be using the correct fits keywords to represent black and white values, or using non-standard keywords.

[ftherrmann]

Let me keep this simple and ask:

What settings to I have to use in order to display luma values of a 32 bit fp fits file (produced by MaximDL) that correspond to the levels I see in a 8 bit 256 value bitmap?

I want the "white" stars to max out at 256 and the darks to min out at 0.    8 bits/mean seems like a reasonable setting, but it doesn't work.  I simply want to see the same 256 bitmap values that I see in CCDStack.

Thanks,

Fred

[Carlos Milovic]

If the data is properly scaled in the normalized range, then just changing that readout option should work.
Dirt workaround: use the Rescale process.
Real solution: find out which value MaximDL is using for storing a saturated signal, and for zero signal. Then, configure the fits format preferences with those values.

FYI, PixInsight uses the convention of interpreting floating point values using the normalized range. That means, 0.0 is black, and 1.0 is white. Other softwares use floating point to store "fractional" ADUs, i.e. 0.0 is black, and 65535 is white... There is no definition in the FITS format on how to proceed with floating point values, so basically each software does what it wants, and that causes all the incompatibility problems. For storing raw data, 32bits f.p. format is highly discouraged for that reason. Unsigned integer values should be used, since that is the output of the ADC. And even then, there are problems with the use of the most significant, or the less significant bits...

[Juan Conejero]

Hi Fred,

Little to add to what Carlos has already said. I only want to stress the fact that raw CCD data should *always* be stored in unsigned 16-bit integer format. This way there won't be any compatibility problems. A 32-bit format won't make your raw data "better"; it will just waste twice the required space.

FWIW, 32-bit floating point images generated by Maxim DL store pixel data in the [0,65535] range, that is, the 16-bit unsigned integer range stored as real numbers (or at least this is what we have seen so far). To properly load these images in PixInsight, you can do the following:

- Open the Format Explorer window.

- Double click the FITS format item on the left column.

- On the FITS Format Preferences dialog, Default Floating Point Input Range section, set upper range = 65535.

- Click OK.

Now you can load these files properly because PixInsight knows which values correspond to the black and white points of the numeric range the image has been referred to. However, remember to reset upper range = 1 when you load floating point images generated by PixInsight.

[ftherrmann]

Carlos and Juan,

Thanks very much for the fits file format information.  I didn't know that there is no standard fits format.  As Juan said it really doesn't make sense to store fits in 32b fp but that's always seemed to be the default fits storage option in Maxim.  Hard to split an electron into multiple parts!  And in the end it all boils down to my 8 bit screen display.

Best Regards,

Fred

PS:  PI is great!  Keep up the good work!