Re: [OT] Re: anchoring end points to objects

["D. Stimits" <stimits idcomm com>]

Ross Burton wrote:
> 
> On 23 May 2001 21:36:12 -0600, D. Stimits wrote:
> > James Henstridge wrote:
> > >
> > > On Wed, 23 May 2001, D. Stimits wrote:
> > >
> > > > Despite being named "portable", PNG is not as well supported for many
> > > > readers/viewers as is jpeg. Jpeg quality is also not necessarily worse,
> > > > it has a controllable "lossy" compression, which can be turned off
> > >
> > > Maybe JPEG 2000 supports lossless compression (bug is supported by even
> > > less programs than PNG), but the normal JPEGs are always lossless (quality
> > > factor 100 does not mean lossless -- just big files).  Jpeg usually forms
> > > artifacts along edges, which doesn't look very good for most diagrams.
> >
> > jpeg compression is always "lossy" as far as I know, but the amount of
> > loss decreases as compression does. It is possible to turn off
> > compression (large files) and not have the loss. The artifacts though
> > are not a result of jpeg itself, but how it is rendered. Put in
> > anti-aliasing, and it'll look as good as anything else.
> 
> Sorry for being so off-topic... but last year I did a computer graphics
> course at uni, and JPEG was one of the subjects.  I think what I'm
> saying here is right, but it was a dull course... :-)
> 
> JPEG has two modes: loosy and lossless.  Almost everyone uses the lossy
> mode, but some programs allow the "100% quality" setting to use the
> losseless algorithm.  Actually this "quality" setting is an abstraction
> - there are two independant variables and several options in the
> algorithm - there is a plugin for Photoshop which lets the user get to
> the hard-code JPEG algorithm.

JPEG can be lossless even in lossy mode compression IF compression is
not used to any significant degree. By specifying lossy compression but
that compression amount should be zero, there is no relevance to loss,
since actually compressing is required to cause the loss.

> 
> The artifacts in JPEG are part of the algorithm, not the rendering.  The
> image is split into sections (typically 8x8 - look at a low quality JPEG
> close up one day) and the colour levels are compared to a series of
> patterns - gradients from left to right, blobs in the corners etc etc.
> The image segment is said to be formed out of a sum of these patterns in
> various levels.  This is hard to explain but quite simple really -
> although it does emphasis why JPEG images are very bad a representing
> line drawings.  Also, red is bad in simple JPEG encoders - Most encoders
> quantise the red channel and average the levels.

Most encoders do probably suck at this. But are you talking about true
color images also, when compression is not at a level that actually
compresses (regardless of using a lossy or lossless mode)? It sounds a
bit like paletted modes. JPEG's can be very good quality for
non-paletted images that are not compressed. If at any point in the
chain a JPEG is compressed and later saved without the compression, all
bets are off, since the first compression loses information permanently.
But a big argument for adding JPEG export is that some browsers don't
work correctly with PNG. Regardless of whether a PNG or JPEG is used, it
would be nice to see high quality Dia snapshots with anti-aliasing.

D. Stimits, stimits@idcomm.com

> 
> Sorry again for the OT post!
> 
> Ross
> --
> Ross Burton                     Software Engineer
> OneEighty Software Ltd          Tel: +44 20 8263 2332
> The Lansdowne Building          Fax: +44 20 8263 6314
> 2 Lansdowne Road                r.burton@180sw.com
> Croydon, Surrey CR9 2ER, UK     http://www.180sw.com./
> ====================================================================
> Under the Regulation of Investigatory Powers (RIP) Act 2000 together
> with any and all Regulations in force pursuant to the Act OneEighty
> Software Ltd reserves the right to monitor any or all incoming or
> outgoing communications as provided for under the Act
> 
> _______________________________________________
> Dia-list mailing list
> Dia-list@gnome.org
> http://mail.gnome.org/mailman/listinfo/dia-list