Baked in Look

[PDR]

I see folks say this a lot along with "poor AVCHD implementation etc." but I disagree. I think Canon knew what they were doing and did it right. They say it's poor because of missing B Frames. But from what I understand B Frames are only for lower bit rates to help them become more efficient. If anything Panasonic's is the only poor implementation of AVCHD because at anything under 24Mbps is considered AVCHD (with included B Frames). I guarantee you if Panasonic did have the needed B Frames in the 17MBps codec it would have easily matched the Canons in IQ. Canon got around this by (like you mentioned) trippling the data rate which was the correct thing to do. Their codec does not need B Frames. I think the next step up would be AVC-Intra. I could be wrong about all of this but I believe this to be correct.

I think we're saying the same thing , but I should clarify. When I said "poor h.264 implentation" I meant in the technical sense of compression efficiency.

Why use 2x the data rate when you could get similar quality by using 1/2 the bitrate? This means you could have recorded 2x longer, spend 1/2 as much money on storage, etc... AVCHD streams can go straight to Blu-Ray or AVCHD disc without re-compression and suffering generation losses. The high 5D/7D data rate precludes this possiblity - it exceeds the max transfer rate.

Well one reason is the hardware encoder. It costs more money to supply a chip that can use more advanced features. Chips advanced enough to for higher quality compression require more electricity (lower battery life), and generate more heat (not too good in models that overheat already). This is counterproductive for the design and target market of "DSLRs"

The 5D/7D compression efficiency is "poor" largely because it's limited to the Baseline profile of AVC. It's not only the lack of b-frames that makes the efficiency worse. It lacks features like CABAC entropy coding, 8x8 intra prediction. It uses 1 reference frame instead of 4. These features are present in for example, AVCCAM streams - so you can get away using a lower bitrate and have decent quality (i.e. the compression is more efficient).

The problem with the GH1 is that it uses a similar inefficient encoding profile of the 5D/7D - but without the extra bitrate. This predisposes it to problems in certain situations, that I'm sure everyone is aware of.

If you throw enough bitrate, any compression format can look decent. And I agree, given the tradeoffs , form factor, and target market, that choosing a lower efficiency, higher bitrate solution was a better choice by Canon. But choosing a lower efficiency, lower bitrate solution by Panasonic wasn't so good of a choice.

8-bit video is bad enough for banding. But when you have a bitrate starved / low efficiency scenario, it accentuates the banding because there is less data to smooth the gradations . Dithering is less efficient because it essentially acts like noise and requires a higher bitrate to function . In the absense of bitrate, you get more pronounced quantization and banding. Banding already present in the source footage (baked in) is bad news.

Anyways, a large part of the "baked in look" is due to compression. But as others have mentioned, there are other factors, like in camera processing, settings, etc.. that can reduce the freedom in post

 

[Martti Ekstrand]

I have a question for you Martti- in that example you showed - that is good example how to correct white balance- but (and I've never tried this but I'm curious), Are you telling me that if you shot w/ gh1 settings : Contrast +2, Saturation +2, etc. that in post you'd be able to dial those settings back to a more nuetral, flatter image?? I can't imagine that's possible without some type of loss. That's what my interpretation of being 'baked in' was. perhaps I'm off

As a general rule it's easier to dial up contrast and dial down saturation in post for least loss of quality. But take neither to a extreme anyway. A image with so low contrast that parts that really should be black are dark grey will yield less luminance steps when contrast is brought back in and banding of gradients can be pronounced. And a image so saturated that areas go flat at maximum colour value will risk have a bland look when bringing down saturation.