Patent: Canon RAW Video

[Axilrod]

Yes!  I knew they'd figure something out.  Hope this is what's coming to the CDSLR. 

I really think aside from really high resolution the CDSLR will require a lot of accessories to do everything a C300 does, hopefully the price will reflect this. 

[outsider]

The sensor on the C300 is not a bayer type but more of a foveon style where the each pixel collects all three colors but at different depths. The 3 RGB pixels are stacked on top of each other, hence no interpolation (like the bayer does) to produce the final image.

I believe the sensor has 1920 x 1080 (or somewhere around there) spots, each of which capture RGB color info, which is why there's no way that sensor will physically do more then 2k.

Sorry, but that's just plain wrong.  Canon lists the sensor features as:

• 8.3 MP sensor
• Shallow Depth of Field
• High Sensitivity/Low Noise
• High signal-to-noise ratio
• Reduced rolling shutter artifacts
• Signal processing equivalent to that of 3-chip RGB systems
• 8 MP Bayer array; no pixel interpolation needed

I stand corrected.
http://nofilmschool.com/2011/11/canon-cinema-eos-c300-4k-sensor-outputs-1080p/
In which case I don't understand how there's no interpolating since the pixel will be an average of several other pixels around it, will it not?
Or is a "pixel" considered (in the C300 case) an array of 2 green, one red and one blue pixel?

[seanmcr6]

Could the C300 not be hacked to provide 4k output?

Nope. The limitation on the C300 is the codec chip...it's from the XF300 and it's the limitation. So, without a completely new hardware chip, you are stuck with the current MPEG2, 2K, 4:2:2 output.

[seanmcr6]

Interesting.

http://www.google.com/patents?id=cKbZAAAAEBAJ&zoom=4&dq=raw%20jannard&pg=PA1#v=onepage&q=raw%20jannard&f=false

That IS interesting....reading that looks like RED has a patent on exactly how the C300 works. Of course, I must be wrong cause Canon wouldn't/couldn't do that...

The next 2 years will be very, very interesting for the cinematic capture industry, that's for sure!!

[neuroanatomist]

In which case I don't understand how there's no interpolating since the pixel will be an average of several other pixels around it, will it not?
Or is a "pixel" considered (in the C300 case) an array of 2 green, one red and one blue pixel?

The latter.  Each 'pixel' is the effectively sum of the four adjacent actual pixles in an RGGB cluster (although they're not read out quite that way).

[NormanBates]

this patent seems to be a specific thing for high-fps footage, which would require too much banding with standard RAW movie compression (which RED didn't invent and doesn't own; Cineform RAW, for example, did wavelet RAW compression before RED)

what this Canon thing does is that it records one full frame, then a few subsampled frames, then another full frame, etc - later on, those subsampled frames will be upsampled using additional information from the adjacent full frames (seen from a different perspective: it's like using twixtor, but with the subsampled frames as a very very nice guide)

I certainly don't want that for regular fps footage, but if it's the only way to get 120fps or 240fps, then it's a very nice thing to have

[Edwin Herdman]

Surely it would have made sense to hold off on the C300 so they could build in this new RAW video capabilities so you can output 4k video. oh well, C300 II will be around the corner I guess, Canon are turning into Apple!

Just on the face of it, this 4K RAW stuff would require much more throughput and probably much more memory capacity than the C300 is likely to offer.

It's interesting that it's Canon filing this patent and not the other guys (well, making a small assumption there).  Further evidence Canon is listening to what people have asked for and moving to get on top in the video space.

[Axilrod]

The patent was submitted in July 2010 So hopefully we'll see this implemented in a DSLR this year.