Problem with DBE

[Astrocava]

Something strange.

I put 64 samples to make model, but i got this message saying that there are less than three samples:

http://img405.imageshack.us/img405/8739/errora.jpg

Sergio

[Juan Conejero]

Hi Sergio,

This happens because less than three samples are statistically significant as background samples. The Wr, Wg and Wb fields of the DBE interface represent the statistical weights for the currently selected sample in the red, green and blue channels, respectively. These weights correspond to the probability of a sample to represent the sky background in the image: a weight value of 1 means that the sample is being considered as fully representative of the background; a weight of 0 means that the sample does not represent the background at all.

What is happening in your case is that, even while you've defined much more than three samples, less than three of them have been assigned weights greater than zero (at least for one channel). To fix this problem you must tell DBE to apply more tolerant sample rejection criteria. Try increasing the Tolerance parameter, and perhaps also Shadows Relaxation. Your goal is to convince DBE that your samples are representative of the background.

Let me know how it goes.

[Nocturnal]

Hmm, this is interesting. To be honest the DBE parameters besides sample size are pretty much voodoo to me so I always use default settings.

[Harry page]

Hi

I try and adjust the tolerance settings as low as I can get while still including the data from the samples

I find doing this very usefull particulary when working with faint nebula it helps not to over flatten and have gone as low as .05


Regards Harry

[Astrocava]

Hi Sergio,

This happens because less than three samples are statistically significant as background samples. The Wr, Wg and Wb fields of the DBE interface represent the statistical weights for the currently selected sample in the red, green and blue channels, respectively. These weights correspond to the probability of a sample to represent the sky background in the image: a weight value of 1 means that the sample is being considered as fully representative of the background; a weight of 0 means that the sample does not represent the background at all.

What is happening in your case is that, even while you've defined much more than three samples, less than three of them have been assigned weights greater than zero (at least for one channel). To fix this problem you must tell DBE to apply more tolerant sample rejection criteria. Try increasing the Tolerance parameter, and perhaps also Shadows Relaxation. Your goal is to convince DBE that your samples are representative of the background.

Let me know how it goes.

Thanks Juan. At least this will be useful to understand a bit more the DBE process. As Noctuyrnal said I only control the size of the box.

Sergio

[Astrocava]

Ok. I think I know what happens. The shots were made using my DSLR modified camera, and after calibrating with DSS and integrated with PI, the red curve is far right from the others, so all Wr values from all samples are zero. I need to put together the curves before doing DBE, perhaps with ChannelMath?

Sergio

Edit: ChannelMath didn't solve the problem, even ABE is unable to do it.

Code: [Select]

* Sample box generation and rejection:
Channel #0
Generating 532 background regions: 100%
Number of selected background regions: 508
Extracting background samples: 100%
Number of background samples: 0
*** Error: Insufficient background samples - couldn't build a meaningful background model
Only with tolerance 10 DBE can model the background, but...