Dithering Doubles Your Megapixels

[avandonk]

I processed this data at 6.3k pixels wide rather than the standard 4.3k of the Canon 5DH's sensor. This is a factor of root two or 12.4 MP to 25MP.The tiff image is 150MB.

You do not need to upsize all your images in your stacks, just the reference image and Registar will produce the larger images in memory.

The final median stacked image will be the size of the starting image even if it is not used in the combination stage.

Canon 5DH,Canon 300mm F2.8 at f/5, Hutech LPR filter.Exposures 10X(1m,2m & 4m) at 400 ISO. The mount was dithered by Guidemaster between each exposure with a radius of 45 pixels of the guide camera. The usual HDR method.

No enhancement was used to torture these pixels. Just median stacked.

Large image 6176X4077 pixels 11.5MB

http://d1355990.i49.quadrahosting.com.au/2009_05/cruxlmed.jpg


Bert

[rat156]

Hi Bert,

I was hoping that the dithering and increase in image size might work a bit like drizzling and enhance the resolution of your widefield shots.

I like the picture, but somethings that can be fixed easily detract from it. You need some flats as something is vignetting in your imaging path giving that green to purple gradient from middle to edge. You still have square stars, which means the image is undersampled, I calculate it at 5.63"/pixel. I'm not sure if you can drizzle the undersampled images to get a higher resolution, but it'd be worth a try.

Cheers
Stuart

[avandonk]

Stuart here is the same data at the native resolution. I did not bother about the vignetting it was more an exercise in potential resolution. I just wanted to show the whole of Crux without resorting to gradienxterminator. The FOV is 7x4.7 degrees. The actual focal length of a Canon 300mm F2.8L is 291 mm. Dithering does work as it is equivalent to drizzling. It is all about the Niquist Theorem and the method of sampling not just the sample interval. Fourier also comes to mind.
6MB
http://d1355990.i49.quadrahosting.com.au/cruxnat.jpg

As we all have square pixels I would assume that ultimately at 700% things would look a bit blocky. It sure beats a 200k jpg. Or an image taken with the Hubble presented at a few hundred kilopixels.

Not all of us record a handfull of stars and some tiny bit of esoteric nebulousity on our entire sensor. So we can have the luxury of giving each star lots of pixels. RegiStar tells me there are over 80k of stars on that image. Some would say noise. You be the Judge!

Bert

[rat156]

That's the problem, without sufficient resolution you can't tell the difference between signals and noise. You'll also be missing out on stuff. Call me old-fashioned, but if I'm going to go to the effort of taking the picture, I want it to show all that is possible. I'd rather double the focal length, dither and do a mosaic of the resultant images to get my 80k stars.

BTW I only counted 65,763 stars, so the rest must be noise ;)

Cheers
Stuart

[Alchemy]

actually its 65766... you missed 3 just in the lower RH corner. :lol:

[avandonk]

You guys were counting on a crappy JPG, my count was on a 200MB TIFF!

Bert

[gregbradley]

Another fabulous image Bert.

Greg.