OpenEXR support

B

bmt

New member
hi, i know the contest is over, but maybe it's still the right place to share ideas about RED.

what i'd like to suggest is to consider supporting ILM's OpenEXR ( http://www.openexr.com ) format on the RED cameras.

does it make sense?

- it's an open format supported by most of the high-end post-production applications.

- the 16-bit floating point data type is same as the half data type in NVIDIA's CG language. what it means is the possibility of hardware acceleration of algorithms around the format, and even wider industry support.

- lossless compression algorithms, with 2:1 compression ratios of images with film grain. (and it's no marketing b.sh. i use this format in a regular basis, and it's PIZ algorithm is really a THING..).

- i can imagine a very film-like workflow with this format. the footage would reflect all the decisions the DOP made during the shooting straight out of the cam, but still there would be lots of headroom for grading and correcting in post.

i hope it makes sense...
cheers
mt
 
I'm not sure that would even apply - OpenEXR, as I understand it, is about combining multiple exposures in one file - no way to do that "live" with video effectively (unless locked down, then it is...hey! A still!). OpenEXR is used mostly for film scans or CG rendered imagery, HDR stuff is used in stills photography to combine multiple exposures of static shots.
 
dear Mike,

since OpenEXR represents light values with floating point numbers, over exposured pixels don't clip above value 1.0, but can hold values like 1.1 or 2 and so on.

in my experiences one of the main drawbacks of today's digital image acquisition systems is that if you over-exposure a pixel you loose detail there. So people - where in doubt - under-exposure the image and lalter set the right exposure in post, which practice makes post work a must and not an option, since the straight output of the camera is NOT RIGHT.

my suggestion is to build a special A/D converter which converts the analog signal from the imager to a floating point value instead of an integer. i'm not an engineer so maybe i'm overlooking something - and it's not a real possibility -, but it would be the ultimate FILM LIKE digital solution. White balanced images straight from the camera without the danger of loosing details.

sorry if i wasn't clear last time, but my idea is more about the floating point representation of pixel values, than about this particular file format. OpenEXR just shows that the industry is pushing things that direction.

cheers
bmt
 
bmt,

You correct in your desire for OpenEXR, RAW workflows can truely benefit from this. We've had customers ask for CineForm RAW to OpenEXR, haven't done it yet, but we are getting closer (although you can currently do it through After Effects.) The reason RAW can make great use of HDR formats is RAW is higher dynamic range than other color spaces. In sensor raw the color primaries will range from 0.0 to 1.0, but once the color matrix is applied to convert sensor data to RGB, the primaries can range from negative values to over 2.0. This information is lost (clipped at 0.0 and 1.0) if you convert to any non-float system (i.e. uncompressed RGB, YUV, etc.) In CineForm RAW we keep all options in float until the final output to preserve the dynamic range, yet not everything will except CineForm RAW, making OpenEXR an excellent export format for the HDR output of RAW cameras.
 
thanks David, for your answer,
and i'm happy that OpenEXR is adopted as an intermediate format, by you and your company. but i want to push things a little further..

as far as i know CCDs are converting light into electrical charge, than a control cirquit converts it to voltage and than it DIGITIZES this voltage. this is the point that i'm interested in.

i'm a 3D and compositing artist so i know much more about post-production than about aquisition, but i don't see the reason why this analog signal can not be converted into floating point numbers instead of integers.

my suggestion seems to turn into a question.. :)

is it insane? or unimaginable? or impossible? both You and Mike has much closer contact to professional camera makers than i - since i have none :huh: - so please help me to understand why not this is the way we capture digital images.

thanks
bmt
 
bmt,

there is no need. The sensor say a minimum and maximum value that can be fully represented with integers (with ranges 0 to 4095, eqilivent to 0.0 to 1.0.) It is the post color processing that generates values greater than 1.0. Also there is no difference for analag to float sampling, float is digital presentation with discrete intervals, just like integers, so analog is digitized into a number discrete levels, integers are simply more effecient on storage for the same precision.
 
dear David,
sorry that i'm keeping bothering you! i'm a real newcomer to shooting, but i know what floating point is, and what things it can represent.
maybe it's my hiatus in shooting knowledge and experience is what makes me act so dummy :).
i just thought that my idea could eliminate the practice of underexposure during shooting, and would make DOP's life a bit easier, and could let them concentrate on the middletones where most of the action is happening.
cheers,
bmt
 
The issue is the white clip point in camera is the same whether the data is float point or integer. It is a hardware limitation of the CMOS or CCD sensor -- it has a hard clip point. However, the RAW shooting mode does give extra room for the DP, however converting to correctly saturday RGB can clip that off, that is why a RAW workflow in post is recommended (or OpenEXR.)
 
You must log in or register to reply here.
Back
Top