[theora] Theora Intra (ptalarbvorm) vs. MJPEG

[Basil Mohamed Gohar]

On 05/12/2010 01:52 PM, Andrey Filippov wrote:
> Basil,
>
> Let me (as a camera designer/maker) remind you what I wrote earlier -
> for the video recording standard MJPEG and intra-only Theora are
> sub-optimal. It seems to be more practical to record compressed Bayer
> data from the sensor so the recorded data preserves virtually the same
> information as real "raw", just using less bandwidth/data storage -
> compression is based not just on the human perception (as in most
> image/video compression formats) but on the noise model of the sensor
> (eliminating "garbage" data from the sensor output) . Theora is great
> for distributing video, but (as is) it is not optimal for the video
> recording purposes. Maybe it is possible to make a "flavor" of it
> suitable for compressing raw Bayer pixels similarly to what we were
> doing with JP4 as a modification of the standard JPEG.
>
> Andrey
>
Andrey,

I did not mean to blow off your suggestion at all, and in fact, it makes
a lot of sense.  But, I guess I don't really have the tools available to
me to present much data on that front.  Even with my minimal technical
knowledge of the area, capturing the non-transformed data straight from
the sensor makes a lot of sense.

However, and this is just an honest question on my part, how easy will
it be for a consumer to work with the resulting video?  The advantage of
using Theora Intra is that it is still fully-compliant Theora video, so
anything that claims to support Theora video should be able to play it
without modification.  If efficiency were the sole concern, and not
ease-of-use with resulting video, then we'd have a lot more options
available to us as to where we can go.

So, having said that, what do you think we can do at this point?  Maybe
you can explain a bit more about the processing of the JP4 video that is
produced by your products?  For example, I downloaded and watched some
of those JP4 videos, and, of course, they were not in a final form, but
I also did not know what to do next with them.  Can you give us some
idea about the chain of processing them after they've been captured?  Is
there a way to make this as transparent as possible for end-users?  I
love that we can discuss this idea further.

And to bring up the point, as Theora does show superiority to capturing
details over JPEG, then adapting the JP4 format to use Theora's intra
coding as the codec also seems like a win too.

I'm anxiously awaiting your response!  :)