Southern Cross at Max Resolution for 300mm

[avandonk]

Something Peter Ward said sent me thinking about extracting all the available resolution from my 300mm widefields. The limit is not the lens but the undersampling of the sensor. Started processing last night with a few new ideas. Finished this morning. This image is a HDR of a single field taken with the Canon 5DH and the Canon 300mm F2.8L.

Large image 10MB 7k pixels across.
http://avandonkbl.bigblog.com.au/data/2/13839/image/cruxhires194220081213100230.jpg

I have not used any star reduction or Richardson Lucy enhancement. It is purely the resolution that is there due to the lens. The trick is to stack upsized images. In this case all images were 7k pixels wide 190MB tiffs.

I have also managed to get rid of all gradients.

Thanks Peter!

Bert

[prokyon]

Hi Bert,

very interesting! I have never heard of this method before.
I will try it too. Thank you for posting.

cheers

werner

[avandonk]

Werner you have not heard of it before because I think I just invented it. There is always slight drift and or rotation so the sensor is sampling at slightly different positions so in theory the resolution of a stack of exposures should exceed the resolution of the sensor. The trick is to stack at a higher resolution so this information is not lost.

Bert

[strongmanmike]

I like your interesting fiddles Bert great image.

I pulled it into photoshop and turned down the red and upped the blue and green and it looked very sparkly and natural, great image :thumbsup:

Mike

[avandonk]

Colour was the last thing on my mind after looking at the data for a long time Mike. I will have a fresh look later. Below is a comparison to a DSS1 image.


Bert

[avandonk]

Here is ananimated gif 800k.

http://avandonkbl.bigblog.com.au/data/2/13839/image/cruxdss1196320081214090445.gif



Bert

[tornado33]

Great work there. Interesting idia re upsizing the subs, I must try that with my 300mm lens images
Scott

[iceman]

Beautiful image, Bert. Not often you see the double stars in an image like this.

Nicely done.

[h0ughy]

wow that is an amazing result. Inspirational !!

[avandonk]

Thanks for the comments. I should add the mount was guided by the 100ED at about 680mm FL. So the accuracy of guiding on the sensor with the 300mm lens for a single exposure would have been better than half a pixel! Otherwise this method would not work. The drift and or rotation over many images was only a few pixels on the Canon 5DH sensor. This form of sampling is well understood mathematically as long as some criteria are met.

Just shows we should all have open minds especially to valid criticism.

Here is an image 9k pixels wide. This is a bit over two times the original number of pixels (4368) of the sensor. It was initially processed at 6k pixels. Then upsized to 9k pixels for final slight RL enhancement in IP.

Large image 14.6MB
http://avandonkbl.bigblog.com.au/data/2/13839/image/Crux9k4436920081219092755.jpg


If the stars look a bit square it is because generally the pixels in our sensors are on a rectangular array. Only pixels on a hexagonal array would make the sampling more isotropic. Only Fuji have done this to my knowledge.

I have also adjusted the colour to Mike S's preference (I think!).

Below is a 100% crop of this image.


Bert

[strongmanmike]

Ah, yes...nice and blue :D

Mike
Blue rules :lol:

[gregbradley]

Fabulous shot. Love the main star colours.

Greg.