LRO-NAC image exposures from sfs

[mkbarker]

Hello,

Could someone elaborate on the meaning and derivation of image exposures by sfs for LRO-NAC images? e.g., how are they computed, what units are they given in, what is the "Local exposure" and how/why is it different from the exposure for the whole image?

The manual says to use --compute-exposures-only "once for a big DEM, before using these for small sub-clips without recomputing them." That implies that there is only one "image exposure" associated with each image and that the big DEM should be bigger than the images in order to compute the correct exposure. Is that true? If so, then would the computed exposure depend on the big DEM's resolution? I'm running some tests on some polar NAC images and the computed exposures do seem to change depending on the big DEM's resolution.

Thank you,
Mike

[oleg-alexandrov]

The exposure value for a camera is computed as the average of the image intensity divided by the average reflectance, both computed over the DEM. Later, if the problem is parallelized using ``parallel_sfs``, the exposures are found just once, over the entire DEM, rather than for each tile. For what these mean, see https://stereopipeline.readthedocs.io/en/latest/sfs_usage.html#mathematical-model. (I just added the above blurb there about this.) Indeed, the exposure will depend on DEM resolution, and in fact, on the precise DEM values. In my experiments, the value should not change much if your images have big well-lit portions, and if the DEM is reasonable. The exposure values affect the steepness of the slopes. This is one of those things we did not spend enough time qualifying, and in fact, we suspect that the local slopes are somewhat underestimated by SfS.

…

On Wed, Jun 28, 2023 at 4:11 PM mkbarker ***@***.***> wrote: Hello, Could someone elaborate on the meaning and derivation of image exposures by sfs for LRO-NAC images? e.g., how are they computed, what units are they given in, what is the "Local exposure" and how/why is it different from the exposure for the whole image? The manual says to use --compute-exposures-only "once for a big DEM, before using these for small sub-clips without recomputing them." That implies that there is only one "image exposure" associated with each image and that the big DEM should be bigger than the images in order to compute the correct exposure. Is that true? If so, then would the computed exposure depend on the big DEM's resolution? I'm running some tests on some polar NAC images and the computed exposures do seem to change depending on the big DEM's resolution. Thank you, Mike — Reply to this email directly, view it on GitHub <#405>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAKDU3HM7XFVASQIDCXZGYDXNS22DANCNFSM6AAAAAAZXYL4FM> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[oleg-alexandrov]

To add, the local exposure is the exposure for a given little tile. That one is only used to determine if in the current tile there are valid pixels from the given image, and hence if to use the given image for the given tile. The pre-computed exposure is used for all tiles in the actual computation. On Wed, Jun 28, 2023 at 5:30 PM Oleg Alexandrov ***@***.***> wrote:

…

The exposure value for a camera is computed as the average of the image intensity divided by the average reflectance, both computed over the DEM. Later, if the problem is parallelized using ``parallel_sfs``, the exposures are found just once, over the entire DEM, rather than for each tile. For what these mean, see https://stereopipeline.readthedocs.io/en/latest/sfs_usage.html#mathematical-model. (I just added the above blurb there about this.) Indeed, the exposure will depend on DEM resolution, and in fact, on the precise DEM values. In my experiments, the value should not change much if your images have big well-lit portions, and if the DEM is reasonable. The exposure values affect the steepness of the slopes. This is one of those things we did not spend enough time qualifying, and in fact, we suspect that the local slopes are somewhat underestimated by SfS. On Wed, Jun 28, 2023 at 4:11 PM mkbarker ***@***.***> wrote: > Hello, > > Could someone elaborate on the meaning and derivation of image exposures > by sfs for LRO-NAC images? e.g., how are they computed, what units are they > given in, what is the "Local exposure" and how/why is it different from the > exposure for the whole image? > > The manual says to use --compute-exposures-only "once for a big DEM, > before using these for small sub-clips without recomputing them." That > implies that there is only one "image exposure" associated with each image > and that the big DEM should be bigger than the images in order to compute > the correct exposure. Is that true? If so, then would the computed exposure > depend on the big DEM's resolution? I'm running some tests on some polar > NAC images and the computed exposures do seem to change depending on the > big DEM's resolution. > > Thank you, > Mike > > — > Reply to this email directly, view it on GitHub > <#405>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AAKDU3HM7XFVASQIDCXZGYDXNS22DANCNFSM6AAAAAAZXYL4FM> > . > You are receiving this because you are subscribed to this thread.Message > ID: ***@***.***> >