[ELGIS] POLL: 12 bit JPEG compression in TIFF [SEC=UNCLASSIFIED]

Bruce Bannerman B.Bannerman at bom.gov.au
Sun Sep 4 18:51:55 EDT 2011


Peter,

Sorry to labour this point.

It will depend on how the end user wants to use the data.

If they want the data for analysis, via say an image processing application, then anything that will degrade the data should be discouraged. This includes compression via a lossy format such as JPEG or ecw, MrSID, JPEG2000 etc. Analysts will typically want to work with the numbers for each 'colour' band in the the imagery

If the user is just after the data as a visualisation backdrop then the JPEG compression should be fine.


For an rpm package, if the TIFF - JPEG compression is only available as an option, then this should be fine.

I wouldn't like to see it set up as the default option though.

Bruce Bannerman


On 2/09/11 11:23 PM, "Peter Hopfgartner" <peter.hopfgartner at r3-gis.com> wrote:

In my current tests, image quality is fine. Anyway, when trying to
produce some maps from TIFF Orthophotos, with uncompressed GeoTiffs 10
maps took me 1.6 s, with JPEG-Compression the same maps were done only
after 53 s.

Peter

On 08/15/2011 11:35 AM, Micha Silver wrote:
> Hi Bruce:
>
> On Mon, 2011-08-15 at 09:53 +1000, Bruce Bannerman wrote:
>> Agreed.
>>
>> JPEG is a lossy compression format.
>>
> Yes, but when I tried the JPEG 12 bit compression, I found that pixel
> for pixel, it was almost exactly the same as the original tiff.
>
>
>> People who use tiff, typically want a lossless format.
>>
>> Therefore, I don't see the need for a JPEG compression of TIFF.
>>
>> If I wanted a lossy compression, I'd typically go for a wavelet
>> compression format like ecw.
> True, but I always get stuck with the restrictive licensing of ECW.
> And if jpeg 12bit gives a file size only double of the ECW compression
> ratio, with *nearly* lossless results, it becomes interesting...
>
>> Bruce Bannerman
>>
>>
>> On 14/08/11 1:44 AM, "Peter Hopfgartner"
>> <peter.hopfgartner at r3-gis.com>  wrote:
>>
>>          --------Micha Silver<micha at arava.co.il>  wrote--------
>>
>>          Subject: Re: [ELGIS] POLL: 12 bit JPEG compression in TIFF
>>
>>          Date: 12.08.2011 12:12
>>
>>
>>
>>          >On Fri, 2011-08-12 at 11:50 +0200, Mathieu Baudier wrote:
>>
>>          >>  What about the compression rates of the standard TIFF
>>          compressions?
>>
>>          >>  (Already available)
>>
>>          >
>>
>>          >AFAIK, regular TIFF compression (Packbits, or LZW and the
>>          like) only
>>
>>          >give about 20% compression. JPEG is around 90-95%, ECW and
>>          MrSid even
>>
>>          >more, and JPEG 12-bit looks to be about 85-90%.
>>
>>          >
>>
>>          Most standard TIFF compression schemas are lossless (similar
>>          to gzip compression in PNG). They might give great results
>>          when there are large areas with constant values. They are not
>>          so performant for orthophotos, where, on the other hand, some
>>          data loss might be acceptable.
>>
>>
>>
>>
>>
>>          >>
>>
>>          >>  On 12 Aug 2011 11:36, "Micha Silver"<micha at arava.co.il>
>>          wrote:
>>
>>          >>
>>
>>          >>
>>
>>          >>
>>
>>          >>  On 02/08/2011 12:01, Peter Hopfgartner wrote:
>>
>>          >>  >
>>
>>          >>  >  On 08/02/2011 09:01 AM, Mathieu Baudier wrote:
>>
>>          >>  >>...
>>
>>          >>
>>
>>          >>  Hi Peter:
>>
>>          >>
>>
>>          >>  I did a quick (amateur) test of jpeg 12 this morning and I
>>          must say I
>>
>>          >>  was surprised with the results. I started with a 1.4 GB
>>          tiff ortho
>>
>>          >>  photo, which I also have as an ECW image. The ECW is about
>>          64MB.
>>
>>          >>
>>
>>          >>  I used the OSGeo4W installation of gdal with the libjpeg
>>          that supports
>>
>>          >>  12 bit, and I made two compressed tiffs. The regular jpeg
>>          compression
>>
>>          >>  came down to about 75 MB, a bit larger than the ECW, but
>>          it's quality
>>
>>          >>  was a bit fuzzy with color changes in the pixels. (Of
>>          course ECW is
>>
>>          >>  also lossy, introducing changes in the pixel coloring, but
>>          keeping
>>
>>          >>  "sharpness").
>>
>>          >>
>>
>>          >>  The 12 bit jpeg-compressed tiff came down to about 120 MB,
>>          only twice
>>
>>          >>  the size of the ECW, but it was almost indistinguishable
>>          from the
>>
>>          >>  original tiff! Both in coloring and sharpness. That was
>>          very
>>
>>          >>  impressive, I must say.
>>
>>          >>
>>
>>          >>  I don't know where the jpeg2000 format is going (if
>>          anwhere).  For now
>>
>>          >>  this 12-bit jpeg compression looks like an ideal
>>          alternative to
>>
>>          >>  struggling with the closed ECW format on one hand, or the
>>          unwieldy
>>
>>          >>  file sizes of uncompressed tiff on the other.
>>
>>          >>
>>
>>          >>
>>
>>          >>  Best, Micha
>>
>>          >>
>>
>>          >>
>>
>>          >>
>>
>>          >>
>>
>>          >>  >  >  Peter
>>
>>          >>  >  >  _______________________________________________
>>
>>          >>  >  >  el mailing list
>>
>>          >>  >  >  el at lists.osgeo.org ...
>>
>>          >>  >
>>
>>          >>  >  This mail was received via Mail-SeCure System.
>>
>>          >>  >
>>
>>
>>
>>          Peter
>>
>>
>>
>>          R3 GIS Srl - GmbH
>>
>>          http://www.r3-gis.com
>>
>>
>>
>>          _______________________________________________
>>          el mailing list
>>          el at lists.osgeo.org
>>          http://lists.osgeo.org/mailman/listinfo/el
>>
>>
>>
>> This mail was received via Mail-SeCure System.
>
_______________________________________________
el mailing list
el at lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/el

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/el/attachments/20110905/66658457/attachment.html


More information about the el mailing list