[theora] NHW Image codec - advice

[Raphael Canut]

Hello,

I need advice concerning (quite) blurred images with the NHW codec, as the
results on these images are not good for normal and lower quality settings
(because for these settings the residual coding is on the first
order (256x256) wavelet image and for quite blurred images there is not
enough energy, information in this band), but the results can be good with
the higher -h2 and -h3 quality settings with larger residual coding
(512x256, more information).

My question is, if I detect that an image is quite blurred (in the
preprocessing function, if for example the 4/5th of the pixels have a
segmentation kernel result under 60/70 then the image can be considered as
quite blurred) then I increase quality setting and select now the -h2
setting.Is that important to not respect the original indicated quality
setting (because this -rare- case the result won't be good)? The overhead
in the compressed file won't be too much, as generally a quite blurred
image has a high compression ratio, typically for a 512x512 image, it will
pass from 28Ko at normal setting to 31/32Ko at -h2 setting, just 3/4Ko more.

Could that switch in the quality setting be acceptable?

Many thanks,
Raphael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.xiph.org/pipermail/theora/attachments/20131114/baefb05a/attachment.htm