Vertical Stepping

[David Newman]

Anyway, the png is updated with CineForm with a slight blur on the blue channel only before pulling the matte... honestly, I see a bit more hair detail on the Cineform matte... in any event, knowing how the act of pulling a matte can mess with things, I've uploaded a layered photoshop file with the raw original conversions of the blue screen frame if anyone wants to experiment more. The Raylight version has black and white levels that differ from CineForm, but that's easy to correct for testing...

Jim,

thanks for updating the PNG (http://ftp.datausa.com/imageshoppe/outgoing/HVX_TESTS/STILLS/RAYLIGHT_CINEFORM_BLUESCREEN.png) to offer a fairer comparision, plus some education on the workflow. The black & White point differences you can see are controllable. CineForm decoder will default to CG RGB (black at 0, white at 255), but you can select VS RGB (Video system RGB) which sets the black at 16 and white at 235 (which preserves the headroom on the YUV source data.)

 

[mvb]

honestly, I see a bit more hair detail on the Cineform matte..

It's false detail created by the artifacts in the Panasonic decoder. For example,
examine the woman's skin at 8X magnification with a slight increase in contrast on both decoders:

http://dvfilm.com/raylight/compare2.gif

The Raylight image shows natural detail of skin but the Panasonic decoder
has false detail of mostly horizontal bars.

 

[David Newman]

Marcus,

It would be good to have some real world examples. Remember we are not interested in the chroma smoothing in the RGB space, as we directly convert using the source 4:2:2 YUV data. It would be interesting to know if the YUV representation is somehow deficient in a measurable extent with the Panasonic decoder, as Jim's own experiment shows the keying from a Panasonic/CineForm combination was very good, maybe with the slightly advantage -- we could be perceiving this wrong, but this is business all but perception.

Do you have data that suggested the YUV output of the Panasonic decoder is lacking? Looking for real world cases that would convince users to consider purchasing Raylight along with their CineForm tools.

 

[mvb]

Marcus,

It would be good to have some real world examples. Remember we are not interested in the chroma smoothing in the RGB space, as we directly convert using the source 4:2:2 YUV data. It would be interesting to know if the YUV representation is somehow deficient in a measurable extent with the Panasonic decoder, as Jim's own experiment shows the keying from a Panasonic/CineForm combination was very good, maybe with the slightly advantage -- we could be perceiving this wrong, but this is business all but perception.

Do you have data that suggested the YUV output of the Panasonic decoder is lacking? Looking for real world cases that would convince users to consider purchasing Raylight along with their CineForm tools.

Look at the test frames here, these were shot with a real camera http://dvfilm.com/raylight/raylightRes.htm

Clearly these artifacts happen before the DCT, or in the DCT itself, and so would be present in the YUV data as well.

 

[David Newman]

Yes I see a difference, the Panasonic decoder seems more sharpened enhancing more of the DCT artifacts. So any chance you will are going make a YUV DShow component? The current HDLink tool will convert from nearly any AVI, although going through RGB would result in possibly greater quality loss than using the Panasonic decoder. We also have needs for the Windows x64 market which the P2 Viewer doesn't run, maybe the same issues with Vista. So an upgrade to DShow will result in a shared customer base.

 

[mvb]

Yes I see a difference, the Panasonic decoder seems more sharpened enhancing more of the DCT artifacts. So any chance you will are going make a YUV DShow component? The current HDLink tool will convert from nearly any AVI, although going through RGB would result in possibly greater quality loss than using the Panasonic decoder. We also have needs for the Windows x64 market which the P2 Viewer doesn't run, maybe the same issues with Vista. So an upgrade to DShow will result in a shared customer base.

I will look into it. I've never messed with DirectShow because the API is so complicated. If there are tools you can recommend I'd appreciate it.

 

[David Newman]

Marcus,

Yes the DirectShow API is complex to get started in, but once you have your first filter, extending it and creating more is fairly easy. One of our early employees created our first DirectShow decoder wrapping, since then we are created 50+ other filters. I now use DirectShow for test projects as I love the convience of GraphEdit, yet I wouldn't know how to create one from scratch. I guess start with a sample from the DirectX SDK.

P.S. Put your 60i to 24p algorithms into a DShow filter and we could and the to our buffered RT capture engine with PPro and HDlink.