Personal View site logo
Make sure to join PV Telegram channel! Perfect to keep up with community on your smartphone.
Where's a good explanation of v-log?
  • I think I'm misunderstanding something about the virtues of v-log and flat profiles. If someone could explain this to me, I'd be grateful.

    This is what I think is going on. V-log delivers a very flat image, where the darkest elements are greyish, and the lightest elements are bright, but not completely blown out. This enables color graders to "stretch" those values, and thus, have some play with the midrange gamma. But doesn't this mean that the color palette leaves out a lot of value?

    I don't mean that it leaves out the highs and lows. It seems to me that if you take a narrow range of luminance, and scale it to a larger range, you don't get pixels with in-between values. Say I have a photo with luminance ranges from 0 to 100, and I scale that to 0-200. Thus, each luminance value would be doubled: 1 becomes 2, 3 becomes 6, 42 becomes 82, etc.) There'd be no real increase in the range of luminances. And in some circumstances, you'd get banding if the jump is too big.

    And if this is the case., then why compress the luminances into such a narrow range? Why not scale it to the camera's full luminance range, and preserve as much of the "in -between" luminances as possible? If your camera has a range of, say, 1-1000, then why reduce the image to 300-700?

    As I said, I don't think I'm understanding this correctly. Anyone got a good FAQ?

  • 37 Replies sorted by
  • It's very simple. A camera sensor has a certain dynamic range, i.e. can "see" things from shadows to highlights. The way the sensor sees the world differs from the human eye. To make the sensor data match human vision, a curve/picture profile is applied to the data it has recorded. Standard video curves/profiles, such as Rec709 for HD video, tend to crush the shadows and the highlights. Flat/log profiles reproduce the sensor data so that nothing gets crushed or clipped, typically resulting in a greyish/washed out-image to the human eye. This gives you more information for post-production color/exposure correction which wouldn't have been there with the crushed shadows and highlights of a standard color profile.

  • Log expands the highlight & shadow occupancy of the luminance range (compressing mids). As a result clipping is less harsh and allows for more filmic grading in post.

  • If you want a purely technical description: http://pro-av.panasonic.net/en/varicam/common/pdf/VARICAM_V-Log_V-Gamut.pdf

    But if you want some context, basically V-Log and V-Log L are the same idea as Canon Log or Sony S-Log, though with specific differences. These documents explain it better:

    Standard cameras record into a display-referred color space: the signals are meant to be sent to a monitor for display, without any manipulation. The images are rendered for a certain type of display, and the look is baked in.

    Camera log spaces such as S-Log and V-Log are scene referred. They represent the image in terms of the way the camera's sensor captured it. The look is not baked in. Contrast, saturation, and highlight handling are up to you and your workflow, instead of being determined by the camera and how it was set during shooting. Scene-referred images aren't meant to be displayed directly. A specific transformation is required to convert them to the color space of a display. For most camera log spaces that transformation involves a matrix or a 3D LUT. Standard color correction tools are insufficient to convert those camera log spaces to a display-referred color space, though many people still try to use them, making the job very difficult and producing inaccurate colors.

  • ... ultimately it's actually much easier to get from A to B with V Log. Else why bother.

  • Here's a link to Barry Green's explanation of how the GH4's V-Log-L differs from the Varicam's V-Log tone curve:

    The chart shows 16 stops from -8 to +8, with zero calibrated to 18% gray reflectance. Note how the flat log segment of the curve extends down to about -4, with a rolloff knee down to 0% reflectance in the bottom 4 stops of the shadow range. This is the part of the dynamic range that must be accurately exposed in order to match the non-log segment of the V-log LUT you select.

    The important issue, however, is that green rectangle Barry drew on the chart at stop 4. That's where the GH4's 12-stop dynamic range ends with highlight clipping, 4 stops short of the 16-stop range of the Varicam's V-Log tone curve. In order to match the V-Log curve, he claims you should set exposure for highlight clipping at 79 IRE. That would appear to sacrifice the camera's 80-109 IRE range, for the sake of allowing you to use Varicam V-Log LUT's interchangeably with GH4 V-Log-L footage.

    The vertical calibration on the chart is a 10-bit scale from 0-1024. On a camera with an 8-bit codec like the GH4, you would have a scale from 0-256. That means highlight clipping would be recorded at around 180, with no image details recorded in the range 181-255. At the other end of the scale, the bottom 4 stops of shadow detail would be recorded in the range 32-48, just 4 bits of data to cover 4 stops of dynamic range.

    In short, 8-bit V-Log-L appears to squeeze out more dynamic range by sacrificing tonal range in both highlight and shadow regions.

  • @LPowell

    IN other topic we already have same.

  • @Vitaliy_Kiselev The V-Log chart was missing in the other topic. That's the key piece of evidence.

  • @LPowell

    Here it is same, only members can see this attachment.

    And chart is present in PDF in post above yours.

  • My little prediction about Barry Green's analysis of Matthew Allard's test of V-Log L on the DVX200: the analysis was not correct and their expectations about clipping are wrong. The ISO was set too low to use the entire V-Log L range. Just increase the ISO setting, and the clipping level goes up closer to where you'd expect it to be.

    I don't believe Panasonic would be so ridiculous as to make V-Log L always clips at 79%.

  • Thanks for all the replies. If I'm understanding this correctly, V-log does the following:

    1. It uses a logarithmic luminance scale, thus enabling color correction tools previously restricted to those who worked with actual film.

    2. This scale reduces a slightly wider range of luminances into the workable range, mostly in the higher values, effectively giving the mode two extra stops of exposure. Basically, it "clips" the brightest elements two stops higher than it would otherwise.

    3. If we expose a stop or two more brightly than we normally do, those extra stops shift down and enable us to gain more detail in the midrange and lower ranges.

    Is that more or less right?

    1. It's a pseudo-logarithmic RGB scale. The luma scale takes properties of the RGB scale, since luma is derived from the RGB values.

    2. V-Log L extends the highlight range compared to a standard photo style at the same ISO setting. You can capture brighter highlights without clipping. But just how much it extends and where it clips depends on your ISO setting.

    3. I'm not sure what you mean. But capturing an extended range of highlights doesn't mean the midtones and shadows are the same. To capture more highlights, the sensor gain needs to be turned down. It's sort of like shooting at a lower ISO and then brightening the image in post: the shadows will be noisier. The best exposure for a standard photo style won't be the best exposure in V-Log L. For the same ISO setting, you need to boost the exposure in V-Log L to achieve comparable noise in the shadows and midtones. (or, for the same exposure, you need to boost the ISO setting)

  • To capture more highlights, the sensor gain needs to be turned down. It's sort of like shooting at a lower ISO and then brightening the image in post: the shadows will be noisier. The best exposure for a standard photo style won't be the best exposure in V-Log L. For the same ISO setting, you need to boost the exposure in V-Log L to achieve comparable noise in the shadows and midtones.

    This make no sense. V-Log has nothing to do with sensor, gain or exposure values of the sensor. It is how raw sensor values are transformed into recorded image.

  • How does V-Log L magically extend the highlight range without turning down the gain? Full well capacity doesn't change because of V-Log L. Panasonic was not throwing away tons of highlight range captured by the sensor in the days before V-Log L. Why do you think V-Log L is not available at the same base ISO as standard photo styles? Because for the same ISO setting, it uses a lower sensor gain, and of course there is a minimum gain. I don't like to debate with you. Just think about it.

  • Full well capacity doesn't change because of V-Log L. Panasonic was not throwing away tons of highlight range captured by the sensor in the days before V-Log L.

    Well, as they openly state, they did throw out stuff, just read release notes.

    Why do you think V-Log L is not available at the same base ISO as standard photo styles?

    What exactly is "base ISO" in your understanding?

    Because for the same ISO setting, it uses a lower gain, and of course there is a minimum gain. I don't like to debate with you. Just think about it.

    I think you are utterly confused. It is not a debate, just describing how thing actually works, it is not magic.

    Actual pixel at sensor collects charge as result of photons hitting it. Hence it depend only on exposure time (with same scene, lens and all). Sensor has analog gain (>=1.0) before ADC. As any amplifier is not ideal, it adds some noise (to the heat noise and some other sensor noise). And ADC produce raw values (they are also being corrected to reduce fixed pattern, and other things).

    VLOG or any LOG format is just another function transforming raw values into image. Mathematically it is different, but it is works on same place as all the normal modes.

  • Perform this test, and you will understand the connection between V-Log L and sensor gain:

    Wait and get your hands on V-Log L for the GH4. Shoot an object in V-Log L at ISO 3200, with the exposure adjusted so that the object is just below clipping in the V-Log L output. Switch the photo style to Standard and shoot the same object as a raw photo at ISO 3200, with the same exposure. Use RawDigger or some such tool to analyze the raw image. Will the object be just below clipping in the raw photo? What's the highest ISO setting that will make the object be below clipping in the raw photo?

  • @Brian_Siano

    [With V-Log-L], if we expose a stop or two more brightly than we normally do, those extra stops shift down and enable us to gain more detail in the midrange and lower ranges.

    As with all log profiles intended for use with a calibrated LUT, V-Log-L is designed to be used at exactly one specific exposure - with highlight clipping at 79 IRE. If you overexposure the footage beyond this point, its dynamic range will no longer match the calibration of the LUT and the results will be inaccurate, particularly in the non-logarithmic 4-stops of shadow detail range. Used in this way, V-Log-L becomes a subjective flavoring, like any other built-in profile.

  • Wait and get your hands on V-Log L for the GH4. Shoot an object in V-Log L at ISO 3200, with the exposure adjusted so that the object is just below clipping in the V-Log L output. Shoot the same object as a raw photo at ISO 3200, with the same exposure. Use RawDigger or some such tool to analyze the raw image. Will the object be just below clipping in the raw photo? What's the highest ISO setting that will make the object be below clipping in the raw photo?

    Well, it was not me who stated all this, hence - do the tests.

    And throw out ISO from all this. Just fix ISO (low is preferable due to obvious reasons) and do not touch exposure. Target of this tests is to get math function. You do not need to adjust anything.

  • If you overexposure the footage beyond this point, its dynamic range will no longer match the calibration of the LUT and the results will be inaccurate, particularly in the non-logarithmic 4-stops of shadow detail range. Used in this way, V-Log-L becomes a subjective flavoring, like any other built-in profile.

    One thing I am amazed is how much stuff can be made from extremely simple math thing. If you start adding um smart words.

  • @VK we need to fix this problem. Shooting log rules topic needed.

  • Shooting log rules topic needed.

    LOL

  • Bullet point style. Do this -> then this. Expose to this. Etc. End the IRE rubbish. Just simple rules.

  • Is it possible that when in V log mode, GH4s exposure meter and zebras will be programmed to recalibrate for accurate V log exposure metering as part of the new firmware? I am thinking for example of BMPCC - in camera zebras are calibrated for it's own film log right? and so I expose shy of zebras clipping at 100% for all important highlights and this works fine when there is sufficient light, gives me maximum safe exposure. Why should GH4 V log be so much more difficult to use? Guess time will tell where the zebras need to be set to protect important highlights

  • IRE is not rubbish. It measures luminance values in the decoded video signal which allows us to measure display brightness without referring to bit depth.

  • IRE is not rubbish. It measures luminance values in the decoded video signal which allows us to measure display brightness without referring to bit depth.

    OK, how about using simple percentage in this case (0-100%) ? As all IRE complications came from reality we had with analog signal.

  • On a camera with an 8-bit codec like the GH4, you would have a scale from 0-256. That means highlight clipping would be recorded at around 180, with no image details recorded in the range 181-255.

    It make really no sense, at all.

    As only marketing idea (to make it look dull and super flat) can lead to this. Or some utter failure that Panasonic proponents want to hide under smart looking terms.