lpowel or similar 4:2 color: noticeable difference?

[mikecazzx]

Without throwing numbers at me, how noticeable is the 4:2 color vs whatever comes on the GH1 normally?

In other words, is using a hack that provides MJPEG in 4:2 more important than say using a higher bitrate AVCHD?

I have been reading but I am trying ti understand whats most import to PQ (for me).

 

[Lucas Adamson]

AVCHD will give you the best picture quality/bitrate ratio.

To achieve similar quality, MPEG bitrate must be much higher, as it has a less efficient codec.

The 4:2:2 colour mode is particularly useful for green/blue screen special effects work, where too much colour compression can ruin the effect.

 

[mikecazzx]

Ok, so it sounds like if I am not doing green screen stuff, there is no PQ benefit to MPJEG and 4:2:2, right? So side by side I would not see a huge difference in color quality, right? Now things being equal, if the card can handle a higher MJPEG bitrate, will it outperform AVCHD?

 

[sammysammy]

if some one can answer this for sure..does the hack actually gives us 4:2:2 , i was reading a thread where a few people thought it was not even with the 4:2:2 checked with ptools.thanks

 

[Lucas Adamson]

Mike, you won't notice any difference in real terms. Contouring is a horrible colour compression artefact, but if the bitrate is high enough, it's not there. AVCHD is a very good, efficient codec.

One of the problems with MPEG, besides the huge file sizes, which make editing difficult for some and recordings shorter per GB, is that it often leads to frame skipping caused by the high bitrates and the GH1 not being able to cope with them.

 

[paulmeyer]

Hey Mike, here is some reading material: http://en.wikipedia.org/wiki/Chroma_subsampling if you're interested.

I don't know if you've tried working with AVCHD files before but i suggest you try doing so, firstly because they they are not nearly as widely supported as motion jpeg Quicktime files, so for one almost any application these days can handle the Mjpeg Quicktime whereas very few apps can handle AVCHD, and those that can, so so while chugging away, as the processor is trying to decode the complicated codec.

Then there is the 4:2:2 issue, we recently had to shoot a hand moving against green screen and we shot the first batch of footage on AVCHD, it was unusable. The 4:2:0 compression totally messes up the edges, and it is amplified if there is any form of motion blur or depth of field blur etc. We re-shot using motion jpeg and it keyed perfectly. Here's one ad, we made four: http://www.vimeo.com/16534930

By the way if you want the best "presets" for shooting with you GH1 use Lee Powell's (LPowell) patches, he is one of the people who have the most testing with these codecs, the hack and the GH1.

I use two setups, one with slightly lower bitrate/image quality but will still play back in camera, and a second that has the max quality/bitrate but will not play back in cam, both MJpeg Quicktime.

Mr Powell has, however recently brought out a patch with the highest bitrate AVCHD settings and his screen grabs look amazing. I haven't used it yet, but will definitely give it a try soon.

Hope this helps.

 

[Lpowell]

Without throwing numbers at me, how noticeable is the 4:2:2 color [MJPEG patch versus the normal 4:2:0 color depth of both MJPEG and AVCHD codecs?]

As Lucas pointed out, 4:2:2 color depth is most useful when rotoscoping green-screen special effects. While the MJPEG 4:2:2 patch does work, it uses 25% of the bitrate to encode twice as much chroma information in the vertical dimension. This is bitrate it might have otherwise used to encode finer image details.

Both MJPEG and AVCHD encoders divide the image into separate monochrome (luma) and color (chroma) images. The luma image is sampled at the full resolution of the video mode and contains most of the fine details. Since human vision is less sensitive to chroma details, the chroma image is sampled at half the resolution of the video mode, both horizontally and vertically. This is standard practice for all video codecs, except at the very highest quality profiles.

What I described in the previous paragraph is called 4:2:0 color depth. With 4:2:2 color, chroma is still sampled at half resolution horizontally, but in full resolution vertically. This asymmetrical sampling is actually more effective for interlaced video since you really want to sample every horizontal scanline in each interlaced field. For progressive video, 4:2:2 doesn't make as big a difference since it only improves vertical chroma resolution.

 

[Ben_B]

I have had pretty good results with keying in 4:2:0 AVCHD on the GH13. I was using Primatte Keyer which can compensate for 4:2:0 artifacting. When he moves you can tell something isn't quite right if you're just watching the edge. This is for a scene where a car drives off a cliff and freezes mid-air. Background plate is video taken with camera out on a crane. Can't put up motion video yet, but it looks pretty good. I highly recommend Primatte Keyer. Sorry, sort of off-topic for this thread.