Cinema mode - Why would I want to use it?

[Joker B]

Well I needed to put this matter to rest so I pulled all the strings that I could at NASA Goddard, Adobe and Canon and the definitive answer is:
"The Canon HV20 CineMode is a Gaussian Algorithm applied to the raw CMOS image who's function is to give the impression of a longer dynamic range (thus reduced contrast). Because the algorithm is 'agnostic' is compresses data and with it, detail uniformly across the frame." So I was wrong when I originally contended that Cinemode was a raw mode. It is not. However, upon further questioning, it turns out that TV/AV mode are not raw either, but that they employ an agnostic sharpening algorithm that increases local and overall contrast which has the side-effect of accentuating noise.
I hope that's helpful. My formula henceforth shall be to use TV a fixed 1/30 shutter along with minus settings in the custom menu. The rest shall be adjudicated in POST, glorious POST.

This one post should be in a sticky.

Or (horrors!) on Canon's web site.

[jet]

Having tried cini mode recently in future i shall only use it on sunny outdoor filming, it darkens the picture too much otherwise.

[Halsu]

Having tried cini mode recently in future i shall only use it on sunny outdoor filming, it darkens the picture too much otherwise.

Well, it's the only mode that doesn't overexpose by default... but you can of course do that deliberately with manual exposure (+/-) if you REALLY want to ;-)

[Joker B]

I cross-linked zephyrnoid's post to dvinfo here along with some extra interpretation, that I'll quote myself on here, because (a) these threads on cine mode keep popping up on various forums, and (b) I'm pretty darned sure I'm right.

-----quote begins-----

What that most-likely means is that the high-precision "raw" data is mixed in-camera with a blurred image of the same data before being quantized (to 4:2:2 or 4:2:0 etc). This is an old not-unusual image-processing trick. It's "agnostic" in that it's applied uniformly, it doesn't seek edges or bright regions or anything like that.

You can try it yourself in AE or Photoshop -- mix a pic with a blurred image of the same pic. The precise mixing scheme can vary; I like multiplying, though that requires a gamma correction first: e.g.

final_color = (original_color^x) * (blurred_color^(1-x))

for x ranging from 0 to 1. You can do this in photoshop by copying a layer, blurring it, change the blend mode to "multiply" and use the "levels" command on each of both layers to change the gamma (that is, raise the color value to a power) to (1/x).

Here's a quick over-done sample with x set to 0.5 (gamma 2.0) and the blur very wide. Note that the visible noise is mostly in the original un-blurred layer, so noise seems to get suppressed, too.

Why not do it in post? Well, you can, but the quality of the unblurred, pre-quantized signal in-camera will be better -- especially if they step on the gamma very hard (for a simple 8-bit image, gamma 2 is a pretty strong step).

(edit postscript: probably, what Canon does is use x=0.5 and rearrange the terms -- multiply the un-gamma-corrected images, and then raise the gamma of the result afterwards. Fewer operations and if the image is high-precision you won't notice that part of the quality loss. There is quality loss, in that we are mixing with a gaussian-processed image -- the good part is that you might pick up some detail in the extreme shadow/highlight ranges. So it depends on your subjective definition of "quality")

[LOGOS PATHOS ETHOS]

I was actually against cinemode when I first tried the hv30 some months ago, but then continued trying it and it really works for the look I want... I do not like surgically sharp images, I like soft edges and a neutral starting point to do color grading with more room.

of course you can grade color in any mode, that's not the point... what cinemode does is also deactivate the gain as when you shoot in spotlight mode, but with the advantage of being able to do white balance... ;-)

so , for me, it's pretty much what I need for the look I want now.

please try not to become "crusaders" for either tv mode or cinemode, I don't consider that smart lol

every mode is useful for diff situations, that's why they're there in the first place, to give you options

[C.W.M.]

With newer cameras, like the HF S100, the gain can be turned off. That's why a lot of people are shooting in TV mode. You can still use Neutral or custom image settings, lock the shutter at 1/48 for 24p (or whatever setting you want) and then have manual control of exposure without worrying about the shutter speed changing or gain being added. In Cinemode, it will still add gain as necessary and you can't turn it off. That's about as close as you are going to get to manual control.

[Halsu]

What that most-likely means is that the high-precision "raw" data is mixed in-camera with a blurred image of the same data before being quantized (to 4:2:2 or 4:2:0 etc).

Nope, i'm pretty certain that is not the case. What you perceive as blurring the image is actually just a lack of artificial in-camera sharpening and lower contrast.

When this lower contrast image is compressed to mpeg2, and probably pushed through noise reduction in the process, some actual detail DOES get lost, but only in areas which are of a very low contrast. The noise reduction is likely the same in all modes, but low contrast in cine makes it appear stronger. The rest of the apparent loss is easily compensated with post-production sharpening.

I personally prefer Cinemode's low contrast non-sharpened image: it gives much more to work with in post (brighter midtones / shadows before clipping highlights to white) and no sharpening artifacts (these can be seen as "ringing" or "outlines" on some subjects in other modes).

I prefer to sharpen my images in post, if at all, as then i'm free to choose the best algorithm for the job, as well as exactly dial in the amount of sharpening i use.

[jet]

Well, it's the only mode that doesn't overexpose by default... but you can of course do that deliberately with manual exposure (+/-) if you REALLY want to ;-)

Anything with sky in the background unless its realy bright will be very dark in cini, it softens it a lot a personal taste thing as well, [+/-] is not as imediately available on these cams as prosumer cams and there will be times when changes are almost instantly needed.There are times when i shall use its special look though.

[Halsu]

Anything with sky in the background unless its realy bright will be very dark in cini, it softens it a lot a personal taste thing as well, [+/-] is not as imediately available on these cams as prosumer cams and there will be times when changes are almost instantly needed.There are times when i shall use its special look though.

Well, i more or less always shoot with locked exposure, so +/- IS instantly available to me. And even if it was not, i prefer slight underexposure of the foreground subject over clipping the skies to white.

Also, cine does not soften the image, it just doesn't apply sharpening in-camera.

[cgbier]

How strong are actually the IR/AA filters in front of the sensor? Does anyone know?

[LOGOS PATHOS ETHOS]

no sharpening artifacts (these can be seen as "ringing" or "outlines" on some subjects in other modes).

THIS ALONE is a good reason to use cinemode

one thing should be added though: of course cinemode changes shutter speeds and exposure, but this is the reason why is not suitable for point-and-shoot kinda things, where you have diff ligthing situations all the time; at a constant lighting scenario, cinemode behaves really well, there might be a little gain, but only if you mess around too much with the manual exposure control , which is still available on the joystick when you operate on cinemode

I believe cinemode is good for situations in which both the light and the motion are predictable and therefore you don't have the shutter changing on you like crazy all the time

every mode is good FOR A DIFF KIND OF SITUATION ; I use TV mode also, even AV when needed..

like I said before, I don't marry anything ;-)

[Halsu]

How strong are actually the IR/AA filters in front of the sensor? Does anyone know?

The AA filter (or, OLP, Optical Low Pass filter) should soften the image to distribute any smaller-than-one-pixel detail over surrounding photosites, just enough to avoid aliasing artifacts. How much exactly that is is guesswork, due to many other factors involved, but it's certainly enough to reduce measurable resolution and make the images look a little soft at 100% size.

In addition to this HV20 and other single sensor cameras use bayer pattern, where there's two green photosites for each blue and red one, in a chessboard like pattern. The processed, "debayered", RGB image is said to have roughly 75% of the nominal resolution in practice.

[jet]

Well, i more or less always shoot with locked exposure, so +/- IS instantly available to me. And even if it was not, i prefer slight underexposure of the foreground subject over clipping the skies to white.

Also, cine does not soften the image, it just doesn't apply sharpening in-camera.

Well it does not resolve as many lines in the card tests i have done, anyway apart from the horrid tape hum i like this cam a lot and its rare for two people to truly like the same look.

[Joker B]

Please Halsu, fill us in on other possible meanings of "Gaussian Algorithm" that don't involve blurring! Because that's what Canon says it is. (An alternative scheme might be to mix the sharpened and unsharpened signals -- kind of the same net result, and usually sharpening is accomplished via differencing the base signal against a blurred signal).

I wouldn't be surprised to find out that they also alter the final quantization curve (or lower the contrast of the blurred component), but I'm pretty darned sure that the blurring (a much more narrow blur than in my example, which was deliberately made large to show the effect on a thumbnail) is key to getting the cine-mode "look."

Here's how to emulate the XH A1's "cine" preset #8: "The parameters of that preset are Gamma: Cine 1, Knee: Low, Black: Stretch, Sharpness: -4, Color Matrix: Cine 1, Color Gain: -20, Color Phase: +5, Red-Blue: -5, Green-Red: -5, Blue-Red: +5, and Red-Green: +12. ."

(yup, a custom color matrix is part of the equation)

So,to pursue the question: "what is 'cine' mode?" I guess two valid questions would be: (1) is the cine mode in the XH A1 the same as that in the HV/HF cameras? And (2) does sharpness -4 completely eliminate sharpening, or just reduce its impact?


--

To get back to the OP question though:

Cinema mode - Why would I want to use it?

the best answer is: because you like the way it looks.

If you don't like it, don't use it

[LOGOS PATHOS ETHOS]

--

To get back to the OP question though:the best answer is: because you like the way it looks.

If you don't like it, don't use it

allow me to concur :-)

[Halsu]

Please Halsu, fill us in on other possible meanings of "Gaussian Algorithm" that don't involve blurring! Because that's what Canon says it is.

Where does Canon say so?? If you're referring to Zephyrnoid's post, it was a joke, dude.

As such, none of this matters, but here we go anyway: "Gaussian" only means a "bell curve". I's a function used in a well known blur algorithm, but gaussian curves can be found in many places... including i.e. image's brightness or audio EQ adjustments, the distribution of IQ scores, harmonic oscillators, artifical neural networks... you get the point. Gaussian function can be used in sharpening algorithms too.

(2) does sharpness -4 completely eliminate sharpening, or just reduce its impact?

Usually in pro and semipro cameras, the lowest possible setting means "no sharpening". I haven't used XH-A1, but if -4 is the lowest it goes, then it's probably "no sharpening", otherwise it's "low sharpening".

To get back to the OP question though:the best answer is: because you like the way it looks.

If you don't like it, don't use it

There's a bit more to it than that: as said earlier, it gives more usable dynamic range, so it's better source for color correction than other modes. And the lack of artifical sharpening makes it better source for greenscreen...

[LOGOS PATHOS ETHOS]

There's a bit more to it than that: as said earlier, it gives more usable dynamic range, so it's better source for color correction than other modes. And the lack of artifical sharpening makes it better source for greenscreen...

Yeah, but this pretty much summarizes it pretty neatly dude, I totally agree. Though Cinemode is not perfect, it does give you more room and not having any sharpening artifacts makes it a pleasant image to work with in post.

It's much easier and natural to sharpen things up in post (if needed) than to remove the sharpening artifacts and soften the image up , at least from my own little personal experience with basic color grading

[Joker B]

Where is it stated that Zephyrnoid's post was a "joke"?

Citing test scores and neural networks and such is a red herring. Where in *image processing* does the term "Gaussian" not apply to weightings in blur convolutions? Because image processing is the topic.

[ejolson]

Nope, i'm pretty certain that is not the case. What you perceive as blurring the image is actually just a lack of artificial in-camera sharpening and lower contrast.

Is it possible that certain models/firmware use different sharpening/filtering algorithms for cinemode? For example, not all HV30s have the same firmware:

http://web.canon.jp/imaging/dcp/firm-e/hv30/

Has anyone compared cinemode on the HV20, HV30 and HV40 in terms of sharpening/flitering? How about the HFS10/11?

[Halsu]

Where is it stated that Zephyrnoid's post was a "joke"?

Isn't it pretty obvious from his opening: "...I pulled all the strings that I could at NASA Goddard, Adobe and Canon..."?

As far as i know, neither Adobe or NASA Goddard Space Flight Center are the main authorities on consumer cameras, so it's safe to assume the Canon part is joke too ;-)

Citing test scores and neural networks and such is a red herring. Where in *image processing* does the term "Gaussian" not apply to weightings in blur convolutions? Because image processing is the topic.

The main point was that "gaussian algorithm" very likely has no relevance to HV20 image processing in real life, so i pulled your leg a little... but to get even more sidetracked, you can actually i.e. create a sharpening filter with gaussian function, or adjust the gamma with a gaussian-shaped modification.

It's used a LOT in image processing, for various tasks.

I also assume you think the color matrix mentioned in the XH-A1 settings is somewhat similar to gaussian blur convolution kernel - which it's not. A color matrix is used for color correction, not for blurring.

[Halsu]

Is it possible that certain models/firmware use different sharpening/filtering algorithm

Possible, sure, likely, not.

[Joker B]

Don't worry, I know what a color matrix is. I've been working with real time digital image processing a long time (used to work w/John Galt during Sony's first forays into digital HD, for example -- a thousand bucks a spool for tape, yay!). And so have agencies like NASA and companies like Adobe. It's entirely reasonable to think that they would know the characteristics of "cine" in great detail -- or would at least know the right person over at Canon to telephone (myself, I'd just call Chuck Westphal & have him redirect as needed).

Your citation of Gaussians in sharpening just echoes my earlier comment about taking a difference between source and smoothed signals. The word I did use was "convolution." If one combines filters into a single kernel, they're still "there," just compacted for efficiency. So I think we're in agreement except for the "joke" part!

I'm surprised, though, that no one has dragged out a set of charts to determine if the "cine" color matrix is a straight contrast move, or it it also manipulates the color in other ways, say with a rotation or skew to make the color responses look more like an older color neg stock, say 5247 (more red range, skimpier blue).

Guess it's one of those things where when you reach that level of exactitude, you probably shouldn't be depending on an $1K camera

[ejolson]

Possible, sure, likely, not.

I was uncertain when I saw the the firmware upgrade to change the algorithm for recording DV on the PAL HV30. Could the resolution be measured using a modulation transfer function

http://www.normankoren.com/Tutorials/MTF.html

to tell what the resolution really is in each mode?

[Halsu]

NASA and companies like Adobe. It's entirely reasonable to think that they would know the characteristics of "cine" in great detail -- or would at least know the right person over at Canon to telephone (myself, I'd just call Chuck Westphal & have him redirect as needed).

Well, maybe you could try calling the dude ;-)

So I think we're in agreement except for the "joke" part!

And of course the more relevant part about cine mode processing.

Guess it's one of those things where when you reach that level of exactitude, you probably shouldn't be depending on an $1K camera

Isn't it pretty obvious it does do some color manipulation, if the quote about XH-A1 settings is correct?

Color Phase: +5, Red-Blue: -5, Green-Red: -5, Blue-Red: +5, and Red-Green: +12. ."

...not sure if that's the case either. Might be.

[Halsu]

Could the resolution be measured using a modulation transfer function

http://www.normankoren.com/Tutorials/MTF.html

to tell what the resolution really is in each mode?

It could, to some extent. People have shot test charts and MTF can be derived from those.

The problem is, cinemode's loss of detail isn't uniform: it only affects low contrast small detail, and as such it's not readily visible on charts which are high contrast subjects.

Also, to make sensible comparison, either the cinemode or "other modes" images should be adjusted to match contrast (lower contrast hides detail that's actually there).