Re: CALL for DISCUSSION: eXIf 20170115

On Tue, Jan 31, 2017 at 12:24 PM, John Bowler <
[email protected]> wrote:

> On Tue, Jan 31, 2017 at 4:39 AM, Glenn Randers-Pehrson
> <[email protected]> wrote:
> > I don't understand why this is an argument against compression.
>
> I think as a proponent of compression you have to argue *for* it with
> a compelling argument.  I don't need to argue against it, and I don't
> consider 0.12%, or 2%, best case file size reduction compelling.
>
> > Consider the files at www.simplesystems.org/users/
> glennrp/eXIf/nikon_b700
>
> Ok, some time today.  BTW my samples were genuinely chosen from a very
> large set solely on the basis of pictures that I felt I could release
> to the public domain.  I assume you meant the parent directory; there
> is only one JPEG in the above directory.
>
> > The Exif profile is 24945 bytes.  The JPEG is 385520 bytes, including
> > the Exif profile.  The Exif profile in a ziXf chunk is... hey, the
> > zxIf examples are missing, but it was about 10k.  So in this
> > (typical for me) case, the savings is 15k/385k, or about 4%
>
> Bad math; you used the size of the JPEG file, the PNG files you have
> posted up there are 795569(exIf) and 780078(zxIf) bytes in size, which
> is actually a 0.98% reduction.  I assume you used libpng, the
> compression of the PNG is quite low.  jpegtopng produces PNG files
> which are about 46kbytes smaller than the ones you made, and manages
> to compress the zXIf slightly better as well.
>
> The size of the PNG files is irrelevant to my argument.  I was
talking about JNG.  I haven't posted JNG files because I haven't
implemented eXIf in JNG yet.  But I can easily estimate the result.
The nikon jpg converted to JNG is currently 25k smaller than the
JPG because the Exif is discarded.  If I were to add an uncompressed
eXIf chunk it would be 385k, approximately the same size as the
JPG.  If I were to add a zxIf chunk instead, containing the same
Exif profile, it would have been 370k.


> The image you used is 1080x580 pixels, so the raw data was about
> 626,000 bytes.  This is quite low for a modern camera.  Indeed none of
> my samples are this small; my smallest, produced by a relative old
> mobile phone, was 1024x1795 pixels.  I cannot imagiine someone who is
> going to convert a JPEG to PNG (presumably for editing) would use an
> image that small.
>

I can imagine it. That's about the size an image on Facebook or
other social media site should be.  I know people do post images
straight out of their cameras, but that doesn't mean anything.


>
> Nevertheless if someone does you are still taking about reducing a PNG
> size by 2%, when, in fact, if the libpng defaults were simply set to
> high zlib compression level and Paeth filtering the size reduction
> (using exIf) is 6%, and this scales with the image size.
>

Whatever.  I didn't "pngcrush" these files; my objective was to
create files to test exIf/zxIf.

BTW, this is a 0.5MPixel image.  The spec for that camera (from the
> Nikon web site) is:
>
> "20.2 MP low-light CMOS image sensor for brilliant photo and 4K UHD
> video quality"
>
> The MakerNote is 15kbytes; this is because the sensor is large (well,
> by todays standards normal sized) and, apparently, the camera does not
> reduce the MakerNote when it is used at very low resolution.
>

Whatever.

Glenn

>

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot

_______________________________________________
png-mng-misc mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/png-mng-misc