Image dimensions for cropping should be multiple of 2

[dshean]

Looking at the kh9pc images...

Size is 345581, 21808
Size is 345492, 21800
Size is 344014, 21714
Size is 344925, 21805
Size is 344713, 21822
Size is 343446, 21714

I recommend in the future that all cropping be a multiple of 2, ideally a multiple of 32 or 64, so that image dimensions are clean for all downsampled overviews. This would make sense for for PC, MC and aerial images, and simplifies initial camera model preparation and processing of downsampled images.

[godinlu]

Good point — initially, I even intended to keep the same size for all images.
However, I was told it wasn’t necessary. The idea of using dimensions that are multiples of 2, 32, or 64 could therefore be easily integrated moving forward.

[dshean]

Regarding the constant cropping size for all images - that would be nice, but potentially remove useful information, unless we are doing more careful frame/mark picking on the PC images. Workaround for now is to solve for a unique principal point and focal length for each image, but in the future, room for improvement.

I don't recall seeing documentation of how the cropping was performed for PC images - would be good to add this to the HISTORY project documentation. It looks like the top/bottom edges were manually identified near the center of each strip, and the collimation lines near the top/bottom exposed portion of the film were also removed. This is fine, but we do lose a bit of overlap between adjacent frames from each camera.

I've been revisiting the KH-9 PC film documentation and thinking about options to do this properly. Will require some effort, but could potentially leverage some of the existing fiducial mark code from hipp.