Should we be using Histograms or Waveforms?

farss wrote on 2/9/2015, 5:42 PM

In order to not derail another thread here are two scopes from the same image in Vegas 10:

The top one, the histogram hints that the levels are incorrect and in need of a Levels Fx to get them legal.

The bottom one, the waveform, shows what is going on and that no level correction is required. This confusion arises because of the very nature of a histogram, from Wikipedia: An "image histogram" is a type of histogram that acts as a graphical representation of the tonal distribution in a digital image. It plots the number of pixels for each tonal value. By looking at the histogram for a specific image a viewer will be able to judge the entire tonal distribution at a glance..
That makes a histogram a good tool if you're trying to judge exposure of a well lit scene. Problems arise when trying to use a histogram as a diagnostic tool especially for the Vegas user where the image processing chain is Computer RGB, the FXs can create pixels outside the range of correct levels and because of the nature of a histogram confusion can set in.

Unfortunately Vegas does add its own level of confusion to using the Waveform Monitor with several checkboxes one of which doesn't exist on hardware waveform monitors. Even worse, the default setting of Studio RGB "Off" doesn't help at all. The only help here is that once it's checked it stays checked. Further confusion can be found with the 7.5 IRE Setup checkbox, Setup simply doesn't exist anymore with digital video :(

As I see it the question for the Vegas user comes down to a difficult choice:

1) A tool that requires no setup, has no option that'll give you the wrong answer but can mislead no matter how you use it.

2) A tool that requires some setup, does take some knowledge to use and is the industry standard for evaluating compliance with standards but will never lead you astray once you know how to use it.

Personally my vote goes for 2).
Of all the choices of scopes Vegas offers the one I have never used apart from the very early days is the Histogram. Not only did I soon realise the Histogram can tell fibs seeing gaps in it freaked me out needlessly when trying to adjust an image.

Bob.

Back to post

Comments

farss wrote on 2/10/2015, 2:36 PM

[I]"We always cling to our earlier toys the most. I remember clearly how hard it was to give up thinking in VU terms when I began working in digital audio. The thing both digital video and audio have in common is a hard ceiling, something that is not represented in IRE Waveform or VU metering displays."[/I]

I was never really involved with analogue video and the last time I used a real VU meter was many decades ago. I agree, we all need to move into the digital age and get VU meters and densitometers out of our heads. That's another good reason not to use a histogram, it is the video equivalent of the VU meter, both rely on integration, the waveform monitor doesn't which is why as an instrument it has moved on from the days of composite video and now works with every digital video system up to 12G SDI.

I think what you're really saying is the waveform monitor needs to move on as well, it should be scaled with the actual digital values rather than IRE values. The IRE values of 0% and 100% should be moveable markers, it still makes sense with digital video to know where reference white and black are but there's so many variations on that today it needs to be configurable.

[I]" The guys at Sonic Foundry who came up with this stuff were not slouches."[/I]

There's something messed up when the histogram in Vegas displays something different compared to every other image manipulation program and the one in the users video camera.
Knowing how it's broken is interesting, it must have taken a fair amount of work to figure that out! Still, broken is broken. It's absurd to suggest that there's not something wrong if I and other Vegas users can shoot video using a camera's histograms and then bring the footage to Vegas and using a tool of the same name see a significantly different result.

By comparison if I do lug an external monitor along to a shoot and use its waveform monitor to judge what I'm shooting what I see back in Vegas using the same tool is the same thing, no confusion.

Bob.

videoITguy wrote on 2/10/2015, 2:37 PM

Careful handling of DSLR footage for those who care has been a well documented issue for years with many an NLE including VegasPro. It has been posted in this forum many times, but the perspective depends on who really cares.
Native DSLR can be handled by many different workflows, but the one that I am personally ruling unlawful is native usage on VegasPro timeline even if it were just for transcoding at that position.

wwjd wrote on 2/11/2015, 8:59 AM

I always view all four scopes at once with histo set on LUM only.
vegas scopes are only accurate at PREVIEW: BEST>FULL
but I use them at lower for general referencing.

TeetimeNC wrote on 2/11/2015, 12:03 PM

Never ever bring direct DSLR raw to the timeline - we have talked about this much before.

VideoITguy, I follow this forum pretty regularly but must have missed this discussion. Can you elaborate on what problems DSLR raw to the timeline causes?

/jerry

VidMus wrote on 2/11/2015, 12:06 PM

Here is something that might help.

Do not stare at his sample pictures too long. LOL!

farss wrote on 2/11/2015, 3:51 PM

[I]" Never occurred to me that native DSLR footage would be the issue and it never occurred to me to even to do a search on it."[/i]

Native DSLR footage should not be a problem with Vegas. That said the combination of the camera's picture profiles you're , the GoPro software and Vegas's scopes and a number of random posts you might have read could cause you no end of confusion.
Really this needs a thread of its own or you can contact me directly through my forum profile.

Bob.

Cliff Etzel wrote on 2/11/2015, 4:16 PM

Email Sent Bob - Thanks!

PeterDuke wrote on 2/11/2015, 5:21 PM

"Here is something that might help."

Well it didn't help me!

The first two shots were under and over exposed, deliberately so. There is no way you could expose these shots for the look he was after by gazing at the histogram. You have to look at the image. As he said, the histogram is what you have, it doesn't have a "correct" look.

PeterDuke wrote on 2/11/2015, 5:59 PM

Something I found useful on my first digital camera, was the ability to flash-highlight burnt out parts of an image. Small areas like sparkle on water could be left burnt out, but the petals on a white rose that is the subject of the photo is quite a different matter.

A histogram does not distinguish between many small highlight spots and one big one, nor does it know what is of interest and what is background.

GeeBax wrote on 2/11/2015, 7:43 PM

I would also add to that the fact that it is quite difficult (at least I find it so) to distinguish specific objects in a histogram, whereas I can do so quite easily in a waveform monitor display.

farss wrote on 2/11/2015, 10:22 PM

[I]"I would also add to that the fact that it is quite difficult (at least I find it so) to distinguish specific objects in a histogram, whereas I can do so quite easily in a waveform monitor display. "[/I]

You'll never workout from a histogram where anything is within the image, that's by design of the instrument.

I'm warming up to the Cineroid EVF with it's horizontal and vertical waveform monitors on the side of the image area, they even throw in a vectorscope in the corner. To make it even easier for the cameraman the waveform's colour changes to draw attention to the over exposed areas.

Bob.
.

Marco. wrote on 2/12/2015, 3:37 PM

I'm also very cautious with the use of the histogram and only use it kind of forensically (to analyze banding and such things). I find it could be very misleading in some cases. For example in an hd project of 1920 x 1080 pixels dimension only 4 pixels of same luma value will generate a histogram bar of 10 percent height.

But I don't think there's something broken. The math behind is:

a = amount of pixels of current frame (which should match project dimension) = 2^b

b = ln(a)/ln(2) = 100 % height of histogram bar

So if there is a histogram bar of 50 % height (100/2), the math to find how much pixels this bar represents is:

2^(b/2)

Or for a bar of 10 % height (100/10):

2^(b/10)

This math always worked correctly for me and once you find out what kind of input makes the histogram to plot a linear graph and – otherway round – what kind of graph a linear pixel increase creates, it's pretty straight.