For some unexplained reason when I use DBE i get a resampled image by a factor of 2. I found in the dropdown where you can set resampling and reset it to a factor of 1 but upon reopening PixI it is again 2. How do I reset the resampling factor in DBE to 1 permanently<G>?
Automatic Resampling in DBE
- Thread starter Jharvey
- Start date
C. Sonnenstein
Well-known member
Hello Jack:
Well, this can't be done at the moment, all parameters in DBE (not only Downsample option) will be reset when you inicialize a DBE sesion.
Perhaps in a future we add this possibility, it is a good idea. On the other hand, a specific parameter can be reused by creating an icon with every process, including of course DynamicBackgroundExtraction.
Well, this can't be done at the moment, all parameters in DBE (not only Downsample option) will be reset when you inicialize a DBE sesion.
Perhaps in a future we add this possibility, it is a good idea. On the other hand, a specific parameter can be reused by creating an icon with every process, including of course DynamicBackgroundExtraction.
Carlos Milovic
Well-known member
Hi Jack
The reason is that background models contains only large scale features. So, it is completely safe to generate a downsampled image without losing any details. A later upsampled imagewill generate almost the same result. So, to speed up DBE, the 2.0 factor is the default. It is more efficient to upsample than to interpolate from DBE's algorithm.
By the way, if you use PixelMath to operate with the background image (or Divide), there is no need for it to match your original image in size. It is rescaled internally.
The reason is that background models contains only large scale features. So, it is completely safe to generate a downsampled image without losing any details. A later upsampled imagewill generate almost the same result. So, to speed up DBE, the 2.0 factor is the default. It is more efficient to upsample than to interpolate from DBE's algorithm.
By the way, if you use PixelMath to operate with the background image (or Divide), there is no need for it to match your original image in size. It is rescaled internally.
Nothing to add to what Carlos^2 <g> have said.
I'd just add that most times you can even downsample background models up to 1:4 without problems. They are by definition extremely smooth images that have only large-scale structures.
I'd just add that most times you can even downsample background models up to 1:4 without problems. They are by definition extremely smooth images that have only large-scale structures.