Log vs Gamma processing.

[Mike Brennan]

Fred Meyers recent quotes on Log vs Gamma made me think more about the notion that log is better way to record the output of a ccd camera.

The often quoted notion is that log is good becuase it records more info in the shadow details where the viewer is most sensitive, has always sounded a little too convenient a statement considering I've yet to see a white paper on it form the perspective of recording direct from camera....

 

 

My thoughts....

1/ ccds are good at capturing shadow detail

2/ log trades steps in the highlights to more steps in the shadows but ccd cameras need all the help they can get in retaining detail in the highlights.

3/ Shadow detail as seen by the viewer is much brighter on a screen than in real life... so isn't it a nonesense to compare?

 

 

How many have done a comparison test?

 

Mike Brennan

[Phil Rhodes]

Hi,

 

I am also cautious about log, and on the information I currently have my inclination is to recommend linear storage for anything captured electronically, CCDs being a lot more linear than film. The reason that's usually given for log storage is that it increases highlight luminance resolution, where film handles highlights well, at the expense of shadow luminance resolution, which film handles poorly. This might make sense for film, although I remain to be convinced, but I'm pretty sure it doesn't make sense for video, especially where you might be underexposing video a stop or so with the intention of gamma-ing the midtones back to normality later on.

 

Phil

[David Mullen ASC]

I've been curious as to whether there is any advantage in shooting a very low-contrast image on the F900, and then color-correcting in post using something like Kodak's LUT for DI work, i.e. something that adds contrast to the monitor so you can color-correct without actually throwing out information on the recording you make.

 

Then you end up with a color-corrected but low-contrast original HD master, and then a LUT can be applied by the people making the film-out or by the post house making linear HD masters for HDTV and standard def video displays.

 

This is how I did the HDCAM-SR DI for "Shadowboxer", but in this case, my original was film and I was transferring to 10-bit 4:4:4 log space. It's also sort of how people shooting with the new 4:4:4 HD cameras are working.

 

Don't know if there is any advantage to mimicking this approach when shooting 3:1:1 HDCAM, versus shooting normally and just color-correcting for a good linear HD image on an HD monitor -- and letting the people doing the film-out apply their normal LUT for HD video originals being transferred to 35mm IN.

Edited November 20, 2005 by David Mullen

[Mark Wilson]

I've been curious as to whether there is any advantage in shooting a very low-contrast image on the F900, and then color-correcting in post using something like Kodak's LUT for DI work, i.e. something that adds contrast to the monitor so you can color-correct without actually throwing out information on the recording you make.

Perhaps I'm sticking my neck out here, but when they make a low-con positive contact print for telecine, isn't that usually printed on very fine-grain ultra-slow stock from a correctly-exposed normal-speed neg? (I've never done a job with that sort of budget myself, so I'm no expert:o)

 

I'm pretty sure you can't get the same results by simply under-exposing the original negative or using contrast reducing filters, otherwise everybody would be doing it that way, wouldn't they?

 

Surely if you lower the image contrast on either a film or a video camera, you're just going to bring up all the grain and/or noise when you restore it to the correct level?

[David Mullen ASC]

A camera negative IS a low-contrast original. If you strike an IP off of that, you have an intermediate with similar contrast. If you strike a low-con print off of the negative, you get more contrast compared to an IP or original neg, but not as much contrast as with regular print stock.

 

In other words, you can't really make a copy of a negative that has MORE information (i.e. less contrast) on it than the original. The idea of using an IP for a telecine is just that it is a color-timed positive version of the negative with similar contrast (gamma.) You don't make an IP because you think you can gain even more information than the original negative has.

 

In terms of getting more noise if you create a low-contrast image with an HD camera, it sort of depends on how you create that low-contrast image. For example, if you do it by using more fill light, you don't add more noise to the image. Nor by using Low Contrast filters. Playing around too far with Black Gamma and Master Video Black, yes, the noise may go up in the shadows.

 

Also, adding contrast in post to a low-contrast image in video does not generally INCREASE the noise level.

 

I don't understand the notion of underexposing to achieve less contrast. More highlight information, yes.

 

"I'm pretty sure you can't get the same results by simply under-exposing the original negative or using contrast reducing filters, otherwise everybody would be doing it that way, wouldn't they?"

 

I don't understand the point you're trying to make -- and what any of this has to do with shooting in HD.

[Mark Wilson]

In terms of getting more noise if you create a low-contrast image with an HD camera, it sort of depends on how you create that low-contrast image. For example, if you do it by using more fill light, you don't add more noise to the image. Nor by using Low Contrast filters.

 

I don't get it. If for example your camera has a 10-bit luminance output and the maximum output possible for each pixel was 700 millivolts, that means each digital "step" represents about .7 millivolts. If by one means or another you restricted the output signal to a limited output voltage range, say between 550 and 700 millivolts by lowering the contrast by some means, the output voltage would only be changing over a 150 millivolt range, which is about 210 x .7 milivolt digital "steps", meaning the luminance would only be encoded with somewhat less than eight bits! Wouldn't that introduce quantitization noise when that was "expanded" to the full 10 bits, regardless of where the 210-step "window" lies on the exposure curve?

Edited November 21, 2005 by Hi ho Silver

[Stephen Williams]

meaning the luminance would only be encoded with somewhat less than eight bits! Wouldn't that introduce quantitization noise when that was "expanded" to the full 10 bits, regardless of where the 210-step "window" lies on the exposure curve?

 

Hi,

 

I have always thought along those lines.

 

Stephen

[David Mullen ASC]

I don't think adding contrast in color-correction would be as simplistic as taking "8-bits out of 10-bits" worth of exposure information and re-expanding them back out to 10-bits.

 

Certainly when you do a 10-bit Cineon log scan of film negative, you often have more information than you ultimately use in the final color-corrected image. Same when you transfer a negative to videotape. Starting out with more exposure information and throwing some away in digital color-correction doesn't automatically lead to quantization errors and banding.

 

But my question was a QUESTION, whether there was any advantage to the notion of creating a lower-contrast image on the HDCAM original and letting the final amount of contrast be added by an LUT when doing the film-out or making a video master. I suspect it's not worth it due to the 8-bit recording of HDCAM, the 3:1:1 color, the high compression, etc. -- in other words, it would be better to shoot an image closer to the final image you want to achieve. But this is unlike how people shooting with the Viper or F-950 are working generally, since they can record a 4:4:4 10-bit image.

[Guest Jim Murdoch]

The idea of using an IP for a telecine is just that it is a color-timed positive version of the negative with similar contrast (gamma.)

 

When I went to Telecine school, the reason the gave us for using a low-con positive print was that when you scan off a negative, the brightest highlights in the original scene are represented by the densest black areas, which means that the photosensors tend to be in near-darkness, and so their inherent signal-to-noise ratio is at its worst there. Unfortunately, when the signal is inverted to produce a positive image, the most noise then occurs on the brightest parts of the picture, where tends to maximize its visibility.

 

Using a low-con positive print ensures that at no point are the photocells ever in total darkness which maximizes their linearity.

 

However it would appear that modern CCD telecine's have much lower noise levels, since you don't seem to hear too much about this technique these days. I must confess I've never really thought about this before.

 

...meaning the luminance would only be encoded with somewhat less than eight bits! Wouldn't that introduce quantitization noise when that was "expanded" to the full 10 bits, regardless of where the 210-step "window" lies on the exposure curve?

Which is more or less why camera manufacturers went to 10 or 12 bit ADCs; if you assume that the final viewer output is always going to be 8 bits, that gives you quite a bit of room for post production fiddling with the contrast. In the bad old days of analog video you normally tried to shoot with as much contrast as you could get away with, since you could always turn it down in post, but there was very little scope for cranking it up without introducing noise.

[Michael Most]

The idea of using an IP for a telecine is just that it is a color-timed positive version of the negative with similar contrast (gamma.) You don't make an IP because you think you can gain even more information than the original negative has.

 

While this is certainly true, the reasons for using an IP for a video transfer are many, albeit some are a bit outdated.

 

First, an IP is not the original negative, so for studios that were and still are loathe to release the original negative to a transfer facility, it is the best substitute. Second, negative dirt is always a problem in transfers done directly from negative elements, and with IP you eliminate this issue (positive dirt is much less problematic because being black instead of white, it is less noticeable). Third, the IP almost never has a 1.85 hard matte - and in the "old days," this was an advantage because in many cases, it would eliminate the need to pan and scan vs. using a hard matted print. And finally, IP's are generally made on stocks that are more fine grained than camera stocks. The telecine picks up on this and generally yields a "quieter" picture from IP's than from negatives. As Jim Murdoch mentioned, with negative transfers, there is considerable noise in white areas of the picture (i.e., the densest part of the negative). With IP's, this noise is effectively eliminated.

[David Mullen ASC]

Plus an IP is color-timed, so the telecine transfer color-correction goes a little faster than with negative...

 

My point is that the negative contains all the original scene information, so even if you could copy it to an even lower-contrast film element, you can't get more information than was originally recorded.

 

But as you say, there is an advantage to working with a positive version of the negative in a telecine in terms of noise in highlights.

Edited November 23, 2005 by David Mullen

[Mark Wilson]

And finally, IP's are generally made on stocks that are more fine grained than camera stocks. The telecine picks up on this and generally yields a "quieter" picture from IP's than from negatives.

 

Errrr... but surely the grain from the original neg is going to be faithfully copied onto the IP. It's not going to make any difference to the telecine whether the grain is inherent in the stock or is simply a "photograph" of the camera neg's grain!

 

However, the IP stock can be very fine grained so that it doesn't add any more grain than necessary, since sensitivity is not an issue.

[David Mullen ASC]

Errrr... but surely the grain from the original neg is going to be faithfully copied onto the IP. It's not going to make any difference to the telecine whether the grain is inherent in the stock or is simply a "photograph" of the camera neg's grain!

 

However, the IP stock can be very fine grained so that it doesn't add any more grain than necessary, since sensitivity is not an issue.

 

Yes, there is no reduction in grain (maybe even a slight increase) by copying a negative onto an interpositive, but since you don't have a reversal in density where the highlights are the darkest areas of the frame, as with negative, you get less noise in the transfer, at least, in the highlights. Plus the image has been timed so there are less swings in overall density shot to shot. Hence the "cleaner" look Mike is talking about. Although as telecines improve, the density of highlights in a negative is less of a noise problem.

[Michael Most]

Errrr... but surely the grain from the original neg is going to be faithfully copied onto the IP. It's not going to make any difference to the telecine whether the grain is inherent in the stock or is simply a "photograph" of the camera neg's grain!

 

However, the IP stock can be very fine grained so that it doesn't add any more grain than necessary, since sensitivity is not an issue.

 

The grain from the original negative is part of the IP image, of course. However, the finer grain of the IP stock effectively "breaks up" some of the original negative grain and makes it appear, for all practical purposes, smaller and more random. It's hard to describe to anyone who hasn't sat in a telecine transfer room and transferred hundreds of IP's, but it is quite real and the effect is noticeable.