Why is DBE rejecting almost all sample pixels?

[Nocturnal]

Hi,

I'm trying to eliminate the LP produced gradient in a Helix image. This typically works fine but somehow I'm having trouble convincing DBE to accept pixel values as being 'correct'. I'm illustrating this with a sample in the top right corner which seems properly placed but almost all pixels are rejected. As you can see in the image the resulting image is not satisfactory.

Is the bright nebula in the center messing with it? If so, why? Shouldn't DBE just look at the samples? With images with large targets like M31 we're only sampling the periphery, right?

[David Serrano]

I think it's related to the fact that the weight for the samples (fields Wr, Wg and Wg in the DBE interface) is zero or nearly zero. The weight can be increased by modifying the tolerance and/or shadows relaxation (but this affects all samples) or by activating the "Fixed" checkbox, which makes that sample have a weight of 1 (ie full weight), which I would not advise.

As for the reasons, I would bet it to have something to do with the deviations of each sample from the median of the image, or something like that.

Other "solution" would be to make more than one pass of DBE .

[Nocturnal]

Thanks David for replying. I typically use DBE with default settings, sometimes tweaking the size of the probes if the star field is very dense. I've never noticed this behavior before.

[Carlos Milovic]

I would increase the tolerance parameter. I think that this behavior is normal, and is related to the uneven background, that in this case may be greater than in your previous shots.

[Nocturnal]

I've increased tolerance about as high as it will go. I actually thought this image had an easy background to extract. It seems pretty smooth I placed lots of samples in the top right corner but that part remains 'off'.



Corner.



Sample placement, image overview.

[Carlos Milovic]

This tolerance is not that high Try with 2.0 or 3.0 (this is measured in sigma units, standard deviation). [BTW, can't remember if DBE allowed those values ]. If there is no succes... well, just try a second pass, the iterative scheme.

[Nocturnal]

Ha, I thought tolerance max was 1 I now tried with '2' and that works much better. How odd that such an apparently smooth image requires very different values from the defaults I typically use. The background is still blotchy but I can tweak that further.

Thanks for your help!

[Carlos Milovic]

Every image is a whole new world