Histogram Analysis

[Stephen]

Could someone just run their eye over my thoughts below and see if I'm correct in assuming the following about my recent test shots?

450D/WO FF/Meg72

1. On a 60s sub (f4.8,) the histogram hits the top of the y-axis and appears to be "lopped" off - this is over exposure due to LP?
   
2. On the same sub, the graph is 3/4 the way over to the right - this obviously needs to move to the left, but breaks up when I adjust. This (the position) is also due to the same LP issue?
   
3. The red channel has a double peak (think camel humps) - I have no idea what this means
   

Thanks in advance! (I hate my new house ....... )

[duncan.p]

From my scant knowledge of histograms...

1. Correct

2. Correct

3. Two strong red signals: one from your image source and the other the LP perhaps??

[ollypenrice]

I think Duncan has it. As for the breaking up when you cut back, that seems odd to me. Just cutting back isn't stretching further so the image shouldn't degrade.

However, a big cut back in RGB tends to shift the colour balance so I do it iteratively. Take a bite out of the left hand side in RGB then look channel by channel and cut back as needed to bring the top left of the RGB peaks into line with each other. Then take another RGB bite and repeat. See if that makes any difference.

Real shame about the LP. H alpha, good and narrow? Baader 7Nm or less?

Olly

[jgs001]

I don't believe a peak that is taller than the histogram indicates overexposure/clipping, just a high volume of counts at that brightness level. If the histogram is hard against the right hand side, that's overexposure/clipping.

Have a look at (ok, it's for daylight photography, but the histogram is the histogram whatever you're shooting).

Understanding Histograms

There is a shot of the moon towards the bottom, the peak on Y is well above the top of the graph, but there's nothing overexposed in that image.

[Explorer 190 MN]

I read somewhere to find the correct exposure time you use a lens with the same aperture and point your camera to the sky and take test shots to test for skyfog,light pollution and natural sky glow.

What you are looking for is how much time it takes to get the histograph to 20% up from the left. The double peak maybe created from noise from the CMOS chip calibrate your frame with a dark frame and check agian.. I am not sure the benefits of filters are great if you are running unmodded camera. My 550D is un modded and work great for some objects and my CCD camera works for other although I am only just learning CCD. I dont know what conditions you are shoot under but I use ISO 800 for everthing from 55sec to 2mins no point going over 5 mins dur to dark noise signal. I also read a great article about that you need at least 26mins of exposure time to start getting better signal to noise ratio. BUT having said all that I have also read that with too short an exposure time for your subs then you lose signal with noise. I took a picture of M45 with 1 min subs for 26 mins, stacked with darks,flats and offset frames. I then took a picture of M45 with 4 5 min subs of which only 3 were usable. I then stacked these with only two dark frames. With the 5 min exposure time the histgraph was off the charts but after calibration I found that the final picture had more contrast and detial....

[themos]

Which software are you using to calculate and display histograms?

I ask because I've been told and have good reason to believe that the on-camera histogram and the Canon DPP histogram are not reliable.

[Stephen]

Thanks for the advice everyone. Currently suffering a terrible case of man-flu so probably won't be returning to this until the new year

Will post back then.

[Explorer 190 MN]

I use on camera histograph just to get a rough idea of where to start. I am slowly moving over to CCD astro camera at the moment yet another learning curve.

[Explorer 190 MN]

Ok get well soon..

[ollypenrice]

I don't believe a peak that is taller than the histogram indicates overexposure/clipping, just a high volume of counts at that brightness level. If the histogram is hard against the right hand side, that's overexposure/clipping.

Have a look at (ok, it's for daylight photography, but the histogram is the histogram whatever you're shooting).

Understanding Histograms

There is a shot of the moon towards the bottom, the peak on Y is well above the top of the graph, but there's nothing overexposed in that image.

I'd have thought, though, that if the peak has a flat top it's because the top has been chopped off it. Ie the target has exceeded the dynamic range of the chip at the bright end of the scale. No?

Nice though it is, your link man's moon is not an astronomer's moon (he's getting the earth within the dynamic range as well, of course.) And that histogram has a massively chopped top, just as his moon has large flat areas off the dynamic range.

I don't claim to be right, I'm just thinking aloud. Here's one of my rare lunar forays (don't laugh!! OK you can...) and its histogram. Now on the image I see two overexposed regions and, lo and behold, two histo peaks cropped off. I take these to be the two offending regions.

Olly

[themos]

, that if the peak has a flat top it's because the top has been chopped off it.

peak intensity in a histogram is the extreme right position. If the histo curve drops off gracefully towards the right, you don't have saturation. If it drops off at some significant height, then you do.

The peak of the hist curve just shows you what your most common brightness leve is. If it looks flat, it means that there's a healthy range of brightnesses in the image.

[Explorer 190 MN]

I'd have thought, though, that if the peak has a flat top it's because the top has been chopped off it. Ie the target has exceeded the dynamic range of the chip at the bright end of the scale. No?

Nice though it is, your link man's moon is not an astronomer's moon (he's getting the earth within the dynamic range as well, of course.) And that histogram has a massively chopped top, just as his moon has large flat areas off the dynamic range.

I don't claim to be right, I'm just thinking aloud. Here's one of my rare lunar forays (don't laugh!! OK you can...) and its histogram. Now on the image I see two overexposed regions and, lo and behold, two histo peaks cropped off. I take these to be the two offending regions.

Olly

The left hand side of the histogram is dark pixels

[RobH]

Here's a link to a tutorial I did about understanding the histogram and image stretching.....

pic and description page

Kh3ldar.....can you post an image of your histogram.... a picture paints a thousand words as they say

Hope you feel better soon....starting to feel pretty dodgy myself at the moment :-(

Rob

[Jessun]

Do you mean the two peaks to the left Olly, that's the dark sky.

The up/down scale is just the number of pixels at that particular brightness - here they represent the dark region of the moon pic ie the sky and some crater shadows.

The distribution right/left is the tell tale for what brightness the pixels have.

On the up/down scale I don't believe data can be clipped.

Below is I think a more intuitive way to present a normal histogram: Dark down the bottom, bright at the top. Each bar sticking out to the right to indicate pixel count at that level.

[Stephen]

Here's a link to a tutorial I did about understanding the histogram and image stretching.....

pic and description page

Kh3ldar.....can you post an image of your histogram.... a picture paints a thousand words as they say

Hope you feel better soon....starting to feel pretty dodgy myself at the moment :-(

Rob

I trashed the original data as it wasn't a proper imaging session, I was testing out my new flattener.

Here's my poor attempt at drawing though, this is more or less the curve I had from a 60s sub. The first hump is only in the R channel.

[jgs001]

That doesn't look like any clipping to me, the histogram tails off before getting to the white point, therefore you should be fine with that as is.

Here's an image with clipping/saturation

and it's histogram

the peak top is chopped off in the display, but that's showing the moon surface detail, the spike to the right hand end, if the overexposed directly lit surface. The histogram display is from Aperture