[Sls-mhdc] green book

[Enrico Magli]

Dear Aaron,  all,

Sure, the text about band reordering is something that I pretty much took from a paper as it is. I agree that reordering is out of the standard, and your outline looks good. I can have another pass at it,  removing material and reshaping as you outlined.

Regarding whether it would be less or more effective with a longer/better predictor, I guess one could speculate either way. If the data are more correlated, then the ccsds-123 predictor might take advantage of that very effectively. Indeed, we do not have data to support either claim, so we'd better state very general conclusions :-)

Yes,  writing standards with Latex would be great,  but it looks as unlikely as seeing politicians act for the best interest of their country...

Cheers, 
Enrico

---
Sent from mobile device, sorry for the inevitable typos

Il 20/ott/2014 18:05 "Kiely, Aaron B (332B)" <aaron.b.kiely at jpl.nasa.gov> ha scritto:
>
> Hi Enrico,
>
> Thank you for writing those sections.
>
> The text for the ESA predictor is indeed nice and compact, and I think it's just right for this section.
>
> My opinion on band re-ordering:
>
> The Green Book is primarily intended to be a user's guide for CCSDS-123, and so my concern about the band-reordering material is that the results are based on other compressors.  Because CCSDS-123 typically uses more than one band for prediction, I would guess that band-reordering provides less of a benefit than it does for a compressor like the ESA one that uses a single band (I could be wrong -- I don't have experimental results at my fingertips to confirm this).
>
> I think we can say that:
>
> * Band re-ordering is outside the standard (there is no header mechanism for communicating a different band order for each image).  But the standard makes no requirement that bands are arranged in order of increasing or decreasing wavelength.  From the perspective of the standard, an image with re-ordered bands is simply a different image to be compressed.
>
> * Band reordering has been shown to provide a benefit for other lossless multispectral and hyperspectral image compressors
>     -  However, for some imagers the benefit is small, and using more bands for prediction (larger values of P) might decrease the benefits of band re-ordering
>
> * Finding the optimum band order for a given image is a computationally intensive problem.  However:
>     - A "good" (but not necessarily optimum) ordering could still provide a worthwhile benefit
>     - It's possible one could produce a default re-ordering on the ground, and use this same default throughout a mission phase, to provide a compression benefit with little impact on onboard computations
>
> So I think we can make these general points to provide the reader some idea of the issue and include the citations to indicate where to look to find more information.  (And your current text makes most of these points.)  But I'm reluctant to include tables of band-reordering numerical results for a compressor other than the one that has been standardized.  The natural question is why we didn't include results for CCSDS-123 since that's the subject of the Green Book, and the conclusions that one might reach about band re-ordering for a different compressor might or might not extend to CCSDS-123.
>
> My recollection is that the WG decided to make band re-ordering something outside the standard, because it was thought to be somewhat unlikely that it would often provide enough benefit to be worthwhile (at least assuming a not-too-small value of P, like P=3).  If there is an interesting example where re-ordering makes a difference using CCSDS-123 for P=3, that might be worth including.
>
> Aaron
>
> ________________________________________
> From: sls-mhdc-bounces at mailman.ccsds.org [sls-mhdc-bounces at mailman.ccsds.org] on behalf of Enrico Magli [enrico.magli at polito.it]
> Sent: Thursday, October 16, 2014 9:06 AM
> To: sls-mhdc at mailman.ccsds.org
> Subject: [Sls-mhdc] green book
>
> Dear all,
>
> I have uploaded to CWE a new file (120x2g-2014Oct10_EM) containing some
> text for the ESA predictor and band ordering. The text for ESA predictor
> is compact. The text for band ordering is a little longer, and has the
> following issues:
>
> - it contains results, but I am not sure these results exactly match the
> ESA predictor results reported in other sections (i.e. not sure if they
> have been obtained using the same version of the software)
>
> - the tables are not formatted to look look like the other tables in the
> document.
>
> - literature survey is not thorough.
>
> Before doing extra work on this section, maybe someone else should check
> the content and see if it looks as it was intended to be. By the way,
> the MS Word document gave me a lot of trouble, crashing Word multiple
> times, apparently because of some figures in Sect. 4.
>
> Feel free to modify both sections as appropriate.
>
> Regards,
> Enrico
>
> --
> --------------------------------------------------
> Enrico Magli
> Politecnico di Torino
> Dip. di Elettronica e Telecomunicazioni
> C. Duca degli Abruzzi 24 - 10129 Torino - Italy
> Ph.: +39 011 0904195 - Fax: +39 011 0904099
> enrico.magli at polito.it
> staff.polito.it/enrico.magli
>
> _______________________________________________
> Sls-mhdc mailing list
> Sls-mhdc at mailman.ccsds.org
> http://mailman.ccsds.org/cgi-bin/mailman/listinfo/sls-mhdc