Noise in images

[Les Tilly]

Hi Guys I just spent a few days at one of the North American Imaging Conference the first speaker was Tony Hallas who talked about noise in images

he did a study on Noise in one of his images he took I believe around 40 subs 40 darks not sure how many flats and he said that he thought this image would be great very low noise, But to his surprise he still had noise

so his results showed at around 16 subs the noise did not get any better and in a graph leveled out, he said the only way to improve on the noise is with software with either Noise Ninja or Neat image and he did demonstrate this, made me a believer I thought I would share this with you guys.

he also said that from now on he will only take around 9 darks

Very interesting seminar

he also took 1hour sub as he thought there would be more detail in this sub than taking 8 x 7 1/2 minute subs he also demonstrated this and I thought the 8 x 7 1/2 minute sub looked better with less noise and more chance to get a airplane trail or sat trail with 1 hour sub.

Sometimes they post the seminar on a website if I find it I will share it so you can all check it out

Regards

Les

[Reggie]

Thanks for sharing les, you could've made this post a few days ago before I took 60 darks

I'd be interested in hearing that seminar and the reasoning

[Les Tilly]

I would be interested in seeing it again also as it moved pretty fast hopefully they will add the file to a website

Les

[Aza]

Sounds like a very interesting seminar. Lets hope it does end up online.

[roundycat]

Hi Les, I'm a bit surprised that someone with the knowledge and experience of Tony Hallas would be surprised about noise in his images. Noise is measurement uncertainty and is one of the three things that you cannot escape in life (along with death and taxes).

I would have thought 20 darks would have been better but the problem you always come up against is that of diminishing returns. You might be able to prove that 100 darks will improve the situation but the truth is they will still be adding noise to the picture. Once the picture is nearly finished noise reduction software will probably be more of a benefit than anything else. Once that is done and the black point set to drop most of the remaining noise into the background that is as good as it gets.

I think the long sub versus many shorter subs argument is another red herring. Once again theory proves one thing but practice proves another. If you are troubled with randomly appearing muck like sat and aircraft trails and cosmic rays, and who isn't, then more subs will always be better than less as the bigger noise problems such as these can be eradicated and in the process low level background noise is reduced at the same time.

Dennis

[Darth Takahashi]

This is a tricky subject and by no means simple. The mathematics tell you only part of the story, the rest is down to the equipment used to gather the data. I used to try and take long subs and a minimum of 4/5 darks and bias frames, no flats. Now I just take as many shorter subs as I can with no calibration frames at all. Since the noise should average out over the greater number of sub's.

This is all under the following assumptions;

1. You are imaging with an uncooled camera, such as a DSLR

2. That noise remains reasonably controlled in the shorter sub's

3. The subs are in the range of 3 - 5mins @ ISO400/800

Is seems that the noise in a DSLR spreads as a function of time.

This is my take on this subject.

Neil.