[gdal-dev] GTiff bit shuffle compression feature request

[Even Rouault]

Le 08/12/2023 à 20:23, Laurențiu Nicola via gdal-dev a écrit :
>
> On Fri, Dec 8, 2023, at 21:11, Even Rouault wrote:
>>
>>>
>>>  - GeoTIFF DEFLATE 280 MB
>> Is it with PREDICTOR=3 ?  (If not, then the ZARR datasets should also 
>> be compressed with FILTER=DELTA to have a fair comparison)
>
> I would expect a fair comparison to be between PREDICTOR=3 and 
> FILTER=DELTA and not the other way around,

yes, poor wording of mine. I meant that if using PREDICTOR=3, one should 
compare with FILTER=DELTA. But looking more closely, they are not 
strictly equivalent. PREDICTOR=3 applies the delta as b[0]-a[0], 
b[1]-a[1], b[2]-a[2], b[3]-a[3] where a[0...3] and b[0...3] are seen as 
the 4 byte representation of the float32, whereas FILTER=DELTA does the 
difference b_float - a_float as floating point. This isn't the same...

> but it does seem to level the field significantly:
>
>  - GeoTIFF DEFLATE PREDICTOR=3 255 MB
>  - Zarr DELTA BLOSC zlib NONE 290 MB
>  - Zarr DELTA BLOSC zlib BIT 266 MB
>  - Zarr DELTA BLOSC zlib BYTE 259 MB
>
>>>  - Zarr BLOSC zlib NONE 281 MB
>>>  - Zarr BLOSC zlib BIT 253 MB
>>>  - Zarr BLOSC zlib BYTE 249 MB
>>
>
> So FILTER=DELTA yields worse compression, while GeoTIFF PREDICTOR=3 
> does very well. No need for Zarr! I don't remember PREDICTOR=3 doing 
> so well in the past, so.. no idea :-).
>
> _______________________________________________
> gdal-dev mailing list
> gdal-dev at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/gdal-dev

-- 
http://www.spatialys.com
My software is free, but my time generally not.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/gdal-dev/attachments/20231208/d5d2da89/attachment.htm>