DSLR vs CCD in Registax - Why such a difference

[midnight (Darrin)]

G'day all again.

Well, I am a bit baffled and stuck as to why I got 2 very different results from what should have been a close contest and in favour of the CCD.

I have been giving the DSLR liveview AVI stacking a go and got some pretty good results. I have since managed to get my hands on a Neximage CCD (just the basic model).

Ok - setup. 8"SN with an IR/UV blocking filter + borrowed a 1.25" Barlow. 1st up is the 40D in liveview capturing AVI directly with the 5x option on. 2nd is the Neximage just capturing the AVI.

40D. 1000 frames. Windows media reports as 768x800 20Mbps. Registax processes at 6fps.
Neximage. 1000 frames. Windows media reports as 640x480 55Mbps. Registax process at 25fps.

The big kahuna here is when after both are stacked and the wavelet is applied, well the 40D appears to have a really deep image allowing the wavelet to be stretched heaps. But doing the same thing with the Neximage, and look at the overprocessing.

Can anyone help me explain what's going on here? The CCD from all indications should be better but I can't get anywhere near the finish of the 40D. Is this possibly due to compression (both are USB 2 direct into a Dell laptop - in fact same port). Maybe bit depth?? I used the same jpg compression for uploading.

Image 1 - Neximage single (notice artifacts - ie blocky - in zoom).
Image 2 - 40D live single (artifacts not there)
Image 3 - Neximage stacked after wavelet
Image 4 - 40D stacked after wavelet.

Buggered. Any ideas as to what's going on? By the way, I have only spent 2 nights with the Neximage to trial CCD imaging before I decide to look at some more decent CCDs when my budget allows

Cheers,
Darrin...

[mswhin63 (Malcolm)]

Hi Darrin

I must admit, I dont have much XP on Nextimage but play around with gain settings to reduce artifacts but this may not be the issue.

First thing with Registax and planetary I use align with "Centre of Gravity" setting and "User Align" for about 130-150 for your magnifaction. With variations in each frame due to atmospherics regular alignment gets confused with fluctating atmospherics (not the case for moon shots). This may be causing your problem but not sure.

Personal preference for me on Wavelets I use "Gaussian - Linear" and increase the "Step Increment" to 1. These setting are really play around stuff to get the best results.

Unrelated to a point, for DSLR, stack no more than 500 images. Unfortunately the frame rate is too slow for Jupiter as the planets rotation is too fast and you will get bluring as the planet rotates.
CCD though I try to get minimum 2000 and select and Limit to 200-300. On Registax this is done setting the "Lowest quality" setting and looking at the slider in the middle for the number of frames to be processed. This ensures to get rid rid of some really bad out of focus frames.

All said and done, I use AVIStack now, it seems to get the best processing. I still use Registax as the "Colour Balance", "Colour Align" and "Histogram" feature are really good, especially if you dont get the white-balance correct.

Good luck with it.

[bmitchell82 (Brendan)]

Just a little insight into the difference your seeing with the images.

A. the 40d is 10mp camera. lots of pixels = more streeeeeetch
B. the nex imager is about 1mp = no streeeeetch.....

thats all i can say its almost like getting your little handy camera and blowing it up to A1 size

[mswhin63 (Malcolm)]

Hi Brendan,

Works a little different for Live View capture, The APT software takes live view which scales down the resolution to approx 1MP. So realistically the image resolution is the same.

This is a Canon issue not APT.

Actually I did a test just a short time ago and it is under 1 MP for the DSLR.

[bmitchell82 (Brendan)]

how the hell does that work? MP is a function of actual pixels i thought? unless its doing some kick ass binning?

[mswhin63 (Malcolm)]

Originally Posted by bmitchell82

how the hell does that work? MP is a function of actual pixels i thought? unless its doing some kick ass binning?

I dont know how, but it would be logical as live view is only used for indication of capture more than photography. It is that APT decided to use it as a way of using it a bit like a webcam instead of a high quality CMOS imager. The real benefit is that is has a better pixel alignment for RGB compared to a webcam so colour alignment is not critical on the final process.

Hopefully I am gonna try some testing tonight as well to check it, I know that when I use 5 x zoom on live view the reult is under 1mp I believe the result is the same for x 1 live view.

[midnight (Darrin)]

A couple of things need to be noted here.

1. The image from the 40D is reported as 768x800 and at 100% "zoom" on the screen, it is physically not much bigger than the Neximage.
2. the image from the Neximage is reported as 640x480.

So I don't see the stretching as the main problem as all I am looking at is the resolution in pixels as reported.

In the Neximage shot, you can clearly see the "pixellating" similar to mpeg artefacts where there is a distinct blockyness. Now if I look at the 40D's reported 768x800 which ain't much larger than 640x480, there is hardly any pixellation or blockyness. This will help explain the registax artefacts after wavelet processing but the two resolutions are almost identical and hence shouldn't have this large difference.

But have a look at my Registax wavelet comparison. Its exactly the same process and exactly the same wavelet. To me this looks like a bit depth problem in that it may appear as if the Neximage is capturing only 8bit and hence the wavelet is decimating it on processing.

What do you blokes think?? In fact if it does turn out to be 8bit vs 14bit then I would be much more comfortable knowing that it's just a limit of the Neximage.

Hey Brendan, excuse my ignorance but what is binning? I have read only briefly about it.

Cheers,
Darrin...

[midnight (Darrin)]

Righto - I think I know what's going on. Did a bit more reading and found that the Neximage "raw" mode is only 5fps and hence what I am probably seeing is AVI compression which explains the mpeg looking artefacts. I need to get a proper CCD program to capture properly.

But at 5fps, it's way slower than ,my 40D but admittedly, the images from the Neximage seem richer and considerably lower noise than the 40D.

The Neximage is 16bit according to Celestron.

Cheers,
Darrin...

[bmitchell82 (Brendan)]

Binning in basic form is combining the pixels into one so 1x1 binning means 1 pixel by 1 pixel, 2x2 binning is 4 pixels 3x3 9 pixels.