Intesity with the Histogram?

[Gasman]

Hi Guys

Still playing with this excellent software as the night skies over here in the UK have been terrible.

One thought I`ve had is I also use a 16bit Mono ccd which takes images in fits format.

This is a typical jpeg copy of a spectrum I took with the ccd and spectroscope, its actually Arcturus!. Software is used to breakdown the light/dark lines to show absorption and emission lines.

 

 

Now if I move the cursor along the spectrum the display shows the intensity in adu`s (analogue-digital units) and with a 16bit ccd each pixel can have upto 64k adu`s before its saturated, I try to get the exposure to give about 30000 to 40000 adu`s , any greater than 64k then the spectrum is saturated and not ideal for processing so I`m wondering how this can equate to dslr`s and raw files?. 

I gather the only dslr intensity measure you have with BYNIKON is the Histogram so how would a `good` histogram compare with a `saturated` image histogram? does the histogram peak disappear?.

cheers

Steve

[astroman133]

Steve,

 

The X-axis of the histogram spans the full gamut of possible intensity values.  With 0 (black) at the left hand end and 64K (fully saturated or white), in your example, at the right hand end. Now it is not typically possible for the X-axis represent every individual value.  A convenient way to shrink is to scale the intensity down to the range of 0-255.  This only takes up 256 pixels of horizontal space and will easily fit in most apps.

 

The Y-axis is a count of pixels. It is typically scaled to the intensity value with the most pixels. So if their are 5000 pixels with the intensity value of 100 and this is the most frequent intensity, the Y- axis could have a scale of from 0 to 5000.

 

The shape of the histogram is similar to a bell curve where a shoulder that rises up from the Y-axis baseline to a peak and then drops back down to a shoulder again at the baseline as you progress from left to right across the graph.  As long as both shoulders are away from edges of the graph the data is not clipped; either black-clipped or saturated.  So for an astrophotograph, the peak should not disappear, but should be somewhere between 30-60 % of the way from the left edge of the graph.

[Gasman]

Excellent thanks for that Rick. It does complicate things slightly as I now have three intensity peaks to contend with (R,G, where I just had one before .

Steve

[Gasman]

That should read (R,G, no idea why it came up (R,G ??

[Gasman]

Thats really weird but B comes out when in a bracket?

[astroman133]

You can choose to see the luminosity histogram or the RGB histograms via the buttons to the right of the graph.

[s3igell]

Not weird at all...  Most of the "Smiley Face" Emoji are defined as standard Text Character sequences of Colon, Semicolon, Caret, Right-Paren, Left-Paren, and/or Right-Brace...

[s3igell]

I gather the only dslr intensity measure you have with BYNIKON is the Histogram so how would a `good` histogram compare with a `saturated` image histogram? does the histogram peak disappear?

 

 

 

The X-axis of the histogram spans the full gamut of possible intensity values.  With 0 (black) at the left hand end and 64K (fully saturated or white), in your example, at the right hand end. Now it is not typically possible for the X-axis represent every individual value.  A convenient way to shrink is to scale the intensity down to the range of 0-255.  This only takes up 256 pixels of horizontal space and will easily fit in most apps.

 

The Y-axis is a count of pixels. It is typically scaled to the intensity value with the most pixels. So if their are 5000 pixels with the intensity value of 100 and this is the most frequent intensity, the Y- axis could have a scale of from 0 to 5000.

 

The shape of the histogram is similar to a bell curve where a shoulder that rises up from the Y-axis baseline to a peak and then drops back down to a shoulder again at the baseline as you progress from left to right across the graph.  As long as both shoulders are away from edges of the graph the data is not clipped; either black-clipped or saturated.  So for an astrophotograph, the peak should not disappear, but should be somewhere between 30-60 % of the way from the left edge of the graph.

 

One needs to be quite careful to NOT equate the Display Histogram with any "Quantitative" assessment of "Intensity" in forms equivalent to ADUs.

 

Quite simply, the Histogram is generated from a "Default Stretch" that is applied to the Thumbnail JPG that is embedded in the RAW File (at least that was the implementation before Guylain moved to the current DCRAW-based Image Renderer).  The Histogram is still useful to indicate the "Relative Intensity/Exposure", although there may be additional considerations as Nikon and Canon are known to place somewhat different Emphasis on Color Rendering and Saturation for their In-Camera JPG Processing.

[astroman133]

It is quite easy, from a programming standpoint and does not require much code, to create a histogram from the image data that comes from the camera.  I have done it for my personal CCD camera control application.  It generates and displays L, R, G, or B histograms on the fly in under a second.

[Gasman]

So are the sensors in DSLR`s 16 bit?. Actually I`ve just opened a .nef file in Astroart 5 and it does seem to show the intensity in adu`s if I move the cursor across the image exactly the same as if it was a fits image . Would be pretty useful if that same feature could be in BYNIKON!.

regards

Steve

Ps I did wonder about that Emoji

[astroman133]

Steve,

 

I don't know about all DSLR's but many are 14-bit. However the values are usually stored as 16-bit integers with a range of 0-4095. Most CCD cameras are true 16-bit with values from 0-65535.

 

You should create a post with your suggestion in the Feature Suggestion Box Forum.

[s3igell]

Yes. Most current Canons and Nikons are 14-bit Sensors (0-16383), while the older Digic-1 and Digic-2 era were 12-bit (0-4095).  And the individual Pixel Values (well, actually the Bayer Matrix ) being stored in 16-bit Integers.

 

But, that is where the standard "Common Knowledge" about comparing DSLR vs CCD Sensors "Sensitivity" and "Bit-Depth" starts to fall into a number of "Gotchas".

 

1) Most all DSLRs and some CCDs implement a Dark Pedestal - an additional baseline value so that the Pixel Values are NEVER near-zero values that could cause readout "non-linearity" or math "div-0" errors.

2) While DACs might output 16-bit data, many of the CCD Sensors have Full-Well values of only 28k or 40k or so - meaning 14-15 bits of actual data range.  Canon Sensor Full-Well Depths range from 28K-40K, too - with some much higher such as the 5DIII at 70K, the 6d at 80K and the 1DX at nearly 90K (per ClarkVision site - its a shame they haven't updated their graphs with current mid-range Canons and Nikons; and that they drop the data point labels off most of the CCDs included in their graphs).

 

Visit the ClarkVision "Digital Sensor Performance Summary" for a lot of additional information which shows that at the Sensor Performance level CCD and CMOS Sensors have much more in common than "common knowledge" allows.

[s3igell]

Back along the Original Topic of the Thread:

If anyone wishes to read through the Archive of O'Telescope Forum Threads, there are a number of detailed discussions about the BYE/BYN Implementation of the Histogram (and of some substantial future Code Work for Guylain):

 

BYE Histogram question

BYE 3.1.0 Histograms - still from In-Camera JPG

More Complex Histogram Manipulation