[cairo] Fwd: problem with cairo scale
zaheer.mot at gmail.com
Fri Nov 30 03:36:00 PST 2007
thanks for the inputs.. i could figure a way out based on the fact that
scale greater than half gives a good resolution. so, right now i have done
scale multiple times.
e.g. to scale 0.25, scale 0.5 write to surface, take the surface,
scale 0.5and write again.
this is giving good resolution. however this is slower.
On Nov 29, 2007 1:35 PM, Kalle Vahlman <kalle.vahlman at gmail.com> wrote:
> 2007/11/29, zaheer ahmad <zaheer.mot at gmail.com>:
> > iam surprised by such basic issue. i hope iam missing something. i still
> > dont know why a scaled down image looks good in gthumb on desktop vs
> > on a x86 scratchbox. Does it have to do something with graphic drivers
> > supported on desktop.
> As Dan was pointing out, this is a known limitation/bug in cairo, and
> has been for quite some time:
> "Cairo's software has a bi-linear filter. And when downscaling at a
> factor greater than 2 the quality gets increasingly objectionable,
> (and quite soon gets to something I would call horrible) . So unless
> bi-linear means something very different than what cairo is doing,
> it's definitely not sufficient for all downscaling."
> AFAIK the situation has not improved since then, so unless willing to
> implement a better downscaling for cairo, your options are limited to
> using an external library to scale the image data or to "live with
> GThumb likely uses some more sophisticated (and undoubtedly somewhat
> slower) algorithm for it's downscaling, that's why it has far
> superioir results.
> I'm not sure if cairo supports scaling eg. xlib surfaces through
> XRender which could be in turn implemented with the graphics hardware
> to achieve both fast and good results.
> Kalle Vahlman, zuh at iki.fi
> Powered by http://movial.fi
> Interesting stuff at http://syslog.movial.fi
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the cairo