Video Levels for Canon EOS 7D

[Chris Wright]

.601 clips tons of data. If you shoot with beautiful colors and then capture with 5DtoRGB, you are clipping with standard def colors. I prefer adobe 1998(from canon’s menu) so my skies don’t get deleted, but that’s just me.

Where did you get your info that canon uses .601?

https://technicolorsoftware.hostzi.com/

[Keith Pratt]

Chris Wright: “all web browsers use sRGB for video/images”

Safari, the latest Firefox, Chrome and IE9 all have varying levels of built-in colour management; others none at all. But that’s for images, not video. With video it’s the player plug-in or module that manages colour, isn’t it? And in that case its expectation would be Rec.601 or Rec.709, surely? Meaning you’d be least likely to run into issues with one of the two.

Chris Wright: “and again,Y′CbCr is not an absolute color space, it is a way of encoding RGB information.”

Who claimed it was?

[Thomas Worth]

5DtoRGB doesn’t clip anything. It decodes every single value, and at full range per the Canon space. ProRes and DNxHD compression does not clip, but most likely “scales” full range values to broadcast range. DPX output doesn’t do any of this, of course.

Canon is using the 601 space, and the way this was determined is as follows:

1. QuickTime clips recorded by the camera are flagged with a “colr” atom of type nclc. This atom tells a program how to deal with the video’s color space. The atom in Canon files specifies a 601 color matrix, so even if this was wrong the opening application would decode 601 anyway. The three values in colr are:

Type: nclc
Primaries: 1 (these are actually 709)
Transfer Function: 1 (again, 709)
Matrix: 6 (601 matrix)

Why the primaries and transfer function are 709, but the matrix is 601, is up to Canon to answer. Fortunately this is all accounted for in the QuickTime file, so apps decode it properly.

2. I have tested (and you can too) decoding Canon video with Rec.709, and the colors are most certainly off. Reds tend to look more like orange. If you decode with the 601 matrix, the colors look correct.

[Chris Wright]

Thanks guys, I was stubborn so I checked it out myself. The crazy canon engineers actually compress the high def signal into SD color(601 matrix) I am going to request a feature change for the next camera model.

https://technicolorsoftware.hostzi.com/

[Thomas Worth]

I’m just speculating here, but perhaps the reason this was done due to image processor limitations. As you probably know, JPEG is 4:2:2 with a 601 color matrix per the JPEG/JFIF spec. We already know the camera writes JPEG files, so it may just apply the same matrix function to video. If so, this would explain the use of the 601 matrix when it’s obviously wrong.

[Guillaume Chadaillac]

Hi guys!
Obviously this thread is a bit old but i m glad i found it!
Finally some people seem to share the same problem that I have! 🙂

I am shooting with a 7d. Encoding in prores LT for editing and color correcting in final cut. I am using a matrox box to do my color correction on a Sony Bravia 47′.

So far so good.

Everything will go on the web, On my personal website which uses jw player in flash and html5.

But I encounter 2 problems:

1) Encoding from prores to h264 using, squeeze, mpegstreamclip, quicktime, yields average results. Colors are not accurate and look washed out. This is not new. The web is plagued with h264 encoding problems but maybe this has to do with a color space problem as well? What do you think? Is there an easy/cunning fix?

2)When videos are online. They obviously look VERY different on all the popular browsers. Firefox and Safari are ok… but chrome … oh my! Mostly supers washed out and merely… off. I made a lot of research with regards to color managed browsers and ICC files… but this seem to only apply to photos … not videos…
Is that right? What can we do during the h264 encoding process that will ensure consistency across browsers?

Thanks a lot guys!!!!