Page 1 of 1
Rc1: Cleaned jpg increased file size.
Posted: Thu Jan 12, 2006 2:48 am
by Taboc
Posted: Thu Jan 12, 2006 10:51 am
by Xyzzy
Yes, there is something wrong about Clean.
Files with no EXIF/IPTC data grow after they are cleaned with EXIF/IPTC related options checked.
X.
Re: Rc1: Cleaned jpg increased file size.
Posted: Fri Jan 13, 2006 3:24 pm
by xnview
Taboc wrote:Some jpg increase file size when Clean:Optimize huffman table is unchecked.
Seems to be correct
All other software with a clean feature have this behaviour...
Re: Rc1: Cleaned jpg increased file size.
Posted: Sat Jan 14, 2006 1:24 am
by Dreamer
xnview wrote:Taboc wrote:Some jpg increase file size when Clean:Optimize huffman table is unchecked.
Seems to be correct
All other software with a clean feature have this behaviour...
So close?
Not always…
Posted: Sat Jan 14, 2006 3:37 am
by Clo
Hello !
Seems to be correct All other softwares with a clean feature have this behaviour…
… that doesn't mean it's normal and logical…
- I tested with the image below, and got a small gain at size, ~4KB less…
—————
- Please, could you check whether you get the same ?
{ BTW : the caption on the pic tells «I've also MY anti-virus
}
Kind regards,
Claude'
Clo
Re: Not always…
Posted: Sat Jan 14, 2006 4:02 am
by FlorianH
Clo wrote:- Please, could you check whether you get the same ?
I can confirm:
a) original file
58.191 bytes
b) original file cleaned with all options except optimize ->
58.191 bytes (same size)
c) original file cleaned with all options (incl. optimize) ->
54.259 bytes
d) file c) cleaned with all options except optimize ->
58.191 bytes strange!
I think from the technical point of view (pierre's view?), behaviour b) is ok. But behaviour d) is not realy easy to explain...
Posted: Sun Jan 15, 2006 12:40 am
by Dreamer
Moved back to New/Open, since it's still not clear if current behavior is standard or not.
Posted: Sun Jan 15, 2006 1:04 pm
by xnview
Dreamer wrote:Moved back to New/Open, since it's still not clear if current behavior is standard or not.
There is no bug here
Posted: Sun Jan 15, 2006 9:55 pm
by Olivier_G
Seems strange that removing data actually increase filesize...
-> can't you actually remove the corresponding metadata space completely ?
Posted: Mon Jan 16, 2006 7:59 am
by xnview
Olivier_G wrote:Seems strange that removing data actually increase filesize...
-> can't you actually remove the corresponding metadata space completely ?
Metadata is removed, but i think that huffmann code is rearranged.
Posted: Mon Jan 16, 2006 12:14 pm
by Xyzzy
xnview wrote:Olivier_G wrote:Seems strange that removing data actually increase filesize...
-> can't you actually remove the corresponding metadata space completely ?
Metadata is removed, but i think that huffmann code is rearranged.
OK, I think this solves the puzzle. But is there any need to NOT optimize Huffman tree?
X.
Posted: Mon Jan 16, 2006 12:52 pm
by Olivier_G
More tests:
I converted a TIFF to JPEG (no Huff Optimization) -> image without metadata: 323644 B
I added several IPTC&Comments -> 323977 B
I cleaned those metadata (no Huff Optimize) -> 323644 B
=> back to original size without metadata (optimize Huff gives 320305 B).
It looks fine to me regarding the way metadata are handled.
(and also regarding Huffman table optimization)
Olivier
Posted: Tue Jan 17, 2006 12:36 pm
by Xyzzy
Xyzzy wrote:
OK, I think this solves the puzzle. But is there any need to NOT optimize Huffman tree?
Self-reply
Yes, when there is image with already optimized coding (fe. by some specialized packer)- applying Huffman could grow it. Looking at Olivier_G test it can be Closed, right?
X.
