[Sls-mhdc] Fall 2017 Draft Meeting Summary (San Antonio)

Wed May 17 19:56:59 UTC 2017

Dear Working Group,

I have produced an updated version of the quantization section (4.8) of  the White Book for CCSDS-123.1-B, available on the CWE at:
	CWE Private/123.1-B/WhiteBook/123x1w-Quantization-2017.05.17.pdf

The new relative error approach actually simplifies the presentation a lot. I think this new version is fairly compact and understandable.

After thinking more about quantization while doing this rewrite, I think we can make some better decisions on a few points. I’d like to see if we can reach consensus on the following items:

(1) I think that the correct way to handle relative error for signed images is to take the absolute value of s^ (as currently done in this newest version of section 4.8) rather than subtract s_min (as done in the most recent draft meeting minutes). We care more about preserving samples with smaller magnitudes, not preserving large magnitude negative samples.

(2) We said at the meeting that relative error control would not be used with periodic fidelity updating. On further reflection, I agree that it may be unlikely that a user would want to use this combination, but I don’t think that it would lead to horrible performance, and allowing it to be used doesn’t really make the standard harder to write or understand. I can’t see a good reason to prohibit this combination.

(3) I’d like to propose that we eliminate Enrico’s action item MHDC-A-1705-7 (to provide results confirming that it’s worthwhile to allow the combination of absolute and relative error). I think that it’s worth allowing this combination, and I think it’s unlikely that Enrico will come up with data to change my opinion on this, in which case the action item is just more work for no real benefit. As Ian pointed out, one might want to be using the relative error limit most of the time, but bound the number of quantizer step sizes to be implemented. I can also imagine that one might want to mostly be using an absolute error limit, but with higher fidelity for rare very dark samples. I think this is something that we can save for a Green Book.

As you review the meeting summary, I hope that you can contemplate these three points and share your opinion to see if we can agree on any or all of them.

Best Regards,
Aaron

> On May 16, 2017, at 4:53 PM, Kiely, Aaron B (332B) <aaron.b.kiely at jpl.nasa.gov> wrote:
> 
> Actually, please instead use version 2, which fixes issues with the equation describing quantization.
> 
> Aaron
> 
> 
> 
>> On May 16, 2017, at 1:38 PM, Kiely, Aaron B (332B) <aaron.b.kiely at jpl.nasa.gov> wrote:
>> 
>> Dear Working Group,
>> 
>> I have uploaded to the CWE a draft meeting summary for your review:
>> 	CWE Private / Meeting Materials / 2017-May / MHDC-MeetingSummary_2017May-v1.docx
>> 
>> Please review and suggest revisions. (Please use track changes, colored text, or some other means of calling attention to your changes to make it easier for me to integrate input from multiple authors.)
>> 
>> I’d like to send out a final version in a week, so I hope that you can respond with suggested changes by Monday May 22 or let me know that you need more time.
>> 
>> Thank you all for an extremely productive meeting.
>> 
>> Best Regards,
>> Aaron
>> 
>> _______________________________________________
>> SLS-MHDC mailing list
>> SLS-MHDC at mailman.ccsds.org
>> https://mailman.ccsds.org/cgi-bin/mailman/listinfo/sls-mhdc
> 
> _______________________________________________
> SLS-MHDC mailing list
> SLS-MHDC at mailman.ccsds.org
> https://mailman.ccsds.org/cgi-bin/mailman/listinfo/sls-mhdc