CSLC quality assurance

[LiangJYu]

This PR adds:

1. computation of min, mean, max of CLSC rasters and corrections
2. determine number of invalid pixels in geocode raster (Very slow!)

To do:

• [x] browse image
• [x] debug slowness or remove feature
• [x] clean up and reorganize

[LiangJYu]

Initial take on CSLC browse image

[vbrancat]

WRT to the browse image, have you interacted with @nemo794? We do want to have the browse image in EPSG:4326 and with a gray color scale. We need to iterate whether we want to use linear or dB scale and other nits ...

[scottstanie]

One initial note that I made for myself awhile ago: something in addition to total number of invalids/nans, what would be great is a check: "Is there a buffer of N invalid pixels around the burst?" This tells us if the bounding bbox we used was big enough. If there's any side where good data runs up to the edge, we have a problem since it probably cut off some data.

This is what I was thinking of making when figuring out what the burst database margin should be... we'll probably need to check a bunch of data at some point to make sure the margins are ok.

[LiangJYu]

Plots linear and log in grayscale. Working on stripping all the unnecessary stuff...

[vbrancat]

@LiangJYu the CSLC-S1 images look cool. Can we improve the contrast by doing some histogram stretching? I have a function to do that if needed

[LiangJYu]

@LiangJYu the CSLC-S1 images look cool. Can we improve the contrast by doing some histogram stretching? I have a function to do that if needed

@vbrancat please share! Thank you!

[LiangJYu]

@LiangJYu the CSLC-S1 images look cool. Can we improve the contrast by doing some histogram stretching? I have a function to do that if needed

Do something like this? found linked from SO here

[scottstanie]

@LiangJYu the CSLC-S1 images look cool. Can we improve the contrast by doing some histogram stretching? I have a function to do that if needed

Do something like this? found linked from SO here

If you want something simpler, I might just try gamma correction + clipping at a percentile threshold. It's only a few lines (e.g. like this function) and can look better than log scaling.

[LiangJYu]

On the subject invalids/nans in image, this is probably best done in the isce3 core module by throwing in an isnan accumulator while loop through all the valid pixels.

[vbrancat]

@LiangJYu ready to be reviewed?

[LiangJYu]

@LiangJYu ready to be reviewed?

In a couple hours. Need to test reorg and add documentation.

[LiangJYu]

Code is untested but barring anything seriously wrong this is pretty close to final implementation.

TODO done by tomorrow noon:

• testing
• more documentation

[vbrancat]

@seongsujeong do you have time to go through this? We need it for the Gamma delivery!