Re: [Openexr-devel] Query: Image Based Lighting and HDRI-enabled cameras

[Brian Willoughby]

What about pixel depth?  Will we ever see 12 bits per component in an HD camera?

I understand that the HD cameras that Lucas used have 10-bit CCDs, but only  
store 8 bits per component on tape due to the compression.  That's why he had  
custom storage hardware built to take the live 10-bit uncompressed outputs and  
store that data without truncating to 8-bit (or compressing in a lossy fashion, 
 
either, I suppose).

I think we have a while before CCDs produce 16-bit half float values (HDRI),  
but we can probably see something beyond 10-bit soon.  I, too, am interested in 
 
finding motion cameras which can capture something akin to HDRI, but it seems  
quite difficult to get such technical details on the higher bit capabilities  
when all the cameras are using MPEG which truncates to 8-bits per component  
before the compression phase begins.  One thing I've noticed about the MPEG-4  
spec (I forget the title of the spec I found) is that it still specifies 8-bits 
 
per component, which truncates higher dynamic range inputs.  This tells me  
that some committees are not aware of the importance of dynamic range.

Brian Willoughby
Sound Consulting


Zap> Anyway. The QUESTION is, is there... or can there be built, a
Zap> camera that works like e.g. the 24p cameras George Lucas used on
Zap> Episode II... but that shot in HDRI *directly*?

You don't want that.  You want raw data from the sensor.  Then you
can convert it to hdri at your leisure, w/o losing anything.

It would be similar to the various raw formats used by the high-end
digital still cameras.

That is not to say lossless(!) compression would be unwelcome, but it
would most likely require an asic or fpga/cpld to keep up.

The current crop of high-end digital still cameras are up to 8 fps
at around 2500 ppi (on the sensor).  At a 24mm by 18mm frame, that
would be on the order of 2560 by 1920.  I've read that 2500 pixels
is enough to prevent the average cinema patron from seeing any
pixelation on the average cinema screen, so speed is the only thing
preventing the use of such sensors.

(The same source said mp film could hit 5000 dpi.  So to truly
duplicate what film can do, in addition to getting the data off
the sensor and into storage faster, you need to quad the pixels
per area and improve the sensitivity to light by a factor of
somewhere between 2 and 8.)

-JimC