Inconsistency in CSIDD Bits Per Pixel

[klucar]

When writing a CSIDD in RGB24I data format, SarPy sets the number of bits per pixel (NBPP) to 8

sarpy/sarpy/io/product/sidd.py

Line 751 in 3354207

basic_args['NBPP'] = 8

However, in the consistency checker, the number of bits it is looking for is 24.

sarpy/sarpy/consistency/sidd_consistency.py

Line 350 in 3354207

sidd_nitf_details.append({'nbpp': 24, 'pvtype': 'INT', 'pixel_type': the_sidd.Display.PixelType})

My thought is that the former needs to be 24 instead of 8.

[bombaci-vsc]

I believe the consistency checker has the error. The NBPP field's full name is "Number of Bits Per Pixel Per Band". 24bit RGB images are stored as three 8bit bands (NBANDS=3, NBPP=8).

The NITF standard is hard to follow, and the exact wording has changed between the MIL-STD-2500C and BIIF documents. However, I think they all restrict NBPP to 8, 16, 32, and 64 for integer pixel types. Here's the 2500C, which may be the most concise:

[pressler-vsc]

Thanks @klucar and @bombaci-vsc; it does like there is a bug here in SarPy.

sarpy.io.product.sidd seems consistent with SIDD Volume 2:

Table 2-6 doesn't really clarify, but the text above it appears consistent:

Given this, I too think the bug is in the consistency checker.

[klucar]

Looks like I guessed wrong! Want me to do a pull request or will this be fixed elsewhere?

[pressler-vsc]

Looks like I guessed wrong! Want me to do a pull request or will this be fixed elsewhere?

I don't believe this is being currently worked and PRs are always welcome! This is a good opportunity/reminder to get at least token sidd_consistency coverage in a unittest.

[klucar]

Here you go!

#545