Autolevel and autocontrast erratic

[gregglee]

I just can't get autolevel or auto contrast to work consistently or really even correctly. Before I take a couple hours to put together examples with images and histograms at multiple settings, I thought I'd just describe it to ask if others are running into this.

I have already turned this in via surveymonkey with example raw files, but that's a bit of a black hole, with no system in place to assign a case number or to report status.

I am a Bibble5 Pro user. But this issue is definitely more than fine tuning from Bibble to ASP. It seems like a bug slipped in during the transition.

With Bibble, autolevels worked most of the time. (meaning no adjustment required) It rarely works with ASP because:

ASP autolevels responds irregularly and unpredictably to clip percentage changes. Starting from .001, then incrementing by .001, nothing happens for one to many increments, than at some point one additional .001 causes a large change. At what value this occurs and the maginitude of the change depend on the image content. Then additional increments produce no further change until the value reaches the next order of magnitude, then eventually another jump change. But by then adjustment is outside the useful range, so just an exercise in checking behavior. Basically ASP autolevel is either on or off with a trigger point and magnitude that varies by image. On the other hand Bibble autolevels responds incrementally to each adjustment to clip percentage, which I see as correct behavior.

So with Bibble I was able to find clip levels to set in camera default that worked most of the time. With ASP the inconsistency makes that impossible.

Auto Contrast (with autolevel off) is a similar situation. Smooth response in Bibble, big jumps in ASP. In addition, in Bibble autolevel and auto contrast produce similar results. In ASP they are rarely similar, with ASP auto contrast frequently throwing the histogram well off the right side compared to autolevels (far to bright in other words). So in ASP auto contrast is rarely a good starting point.

Perhaps related, highlight tool seems more prone to produce the infamous color band artifacts in ASP than Bibble. And this is especially true if highlight and autolevels are used together.

[claudermilk]

IMHO autolevels sucks now and is completely useless for the reasons you have pointed out. I've just stopped using it & manually adjust my endpoints. Hopefully this will be fixed in the next release.

[Tadjio]

I too have been struggling with AutoLevel but initially thought it might be as I am new to Bibble/ASP.
My comparison is with Levels in PaintShop Pro (PSP).

My first reaction is that 0.200 is far too high as a 'default' setting.
Surely there must be a way to set it lower for all images as a starting point?

I experimented and found that 0.002 has no effect at all but 0.003 upwards does.
I find AutoLevel too harsh and aggresive - it seems to be either on or off without any subtlety.

With PSP Levels I can set my Options to 0.01% Threshold for Histogram Clipping and it works very well in Auto mode.
It then displays where the resultant settings are (in the range 0 for Shadows to 255 for Highlights).
For example it might suggest Highlights at 208 and Shadows of 4 whereas
in ASP the same image looks terrible and blown with a setting of 0.2, is mediocre at 0.02 and unchanged at 0.002

All this testing is not helped by the poor handling of Highlights in ASP.
I find that Perfectly Clear is likewise harsh and aggresive.

[afx]

My first reaction is that 0.200 is far too high as a 'default' setting.

Yup, something like 0.02 is much more sensible...

Surely there must be a way to set it lower for all images as a starting point?

This is set just like any other image default...

I experimented and found that 0.002 has no effect at all but 0.003 upwards does.
I find AutoLevel too harsh and aggresive - it seems to be either on or off without any subtlety.

I think the general idea of autolevels is broken by design....

Use autocontrast as a start and go from there.

cheers
afx

[Tadjio]

I think the general idea of autolevels is broken by design....
Use autocontrast as a start and go from there.
cheers
afx

I got quite excited for a minute and tried Auto Contrast.
I couldn't see how it got its parameters and so read the Help system:

• AutoContrast — click this button to set the output white and black points based on image content.
It uses the AutoLevel values (highlight and shadow) to perform a one-time adjustment to increase contrast.
The main difference between this tool and AutoLevel is that the output white and black values can be more easily adjusted than the AutoLevel values.

I screen dumped the Tools Panel of the Curves Histogram for both AutoLevel and AutoContrast and they appear identical in PSP, as verified by doing a 'Subtract'.

AutoContrast vs AutoLevel

AutoContrast.jpg (7.52 KiB) Viewed 8723 times

Back to the black board as it seems that AutoContrast is also "broken by design"

[gregglee]

I think the general idea of autolevels is broken by design....

What does that mean? (Other than you agree that AL is broken.)

Use autocontrast as a start and go from there.

My experience is that auto contrast is more broken than auto levels.
And if you "go from there" (I assume this means adjust the sliders after AC), what is the point of doing an extra AC step? Just adjust the sliders.

[afx]

Back to the black board as it seems that AutoContrast is also "broken by design"

Nope, you did not get the difference.....

Autolevels is continuously adjusting the histogram, so it can interfere with other image adjustments.
Though AS now has it earlier in the pipeline than Bibble, making it less messy, it still is a continuous auto adjustment.
Autocontrast adjusts the histogram only once. And then you can tweak it from there as it conveniently sets the levels handles (the little triangles at the bottom of the curves tool).
This is a huge difference.

And if you "go from there" (I assume this means adjust the sliders after AC), what is the point of doing an extra AC step? Just adjust the sliders.

It is significantly faster. Often enough no further adjustment is needed or only one for the triangle needs to be moved after pressing AC.

The only thing missing is a hotkey for autocontrast....

cheers
afx

[roland65]

Congratulations! You've found a bug in ASP. I tried the same autolevels manipulation and went to the same conclusion : this tool is completely buggy for now!
The question is : how can we report this bug to the dev team? By means of the Bibble forum? Or is there another way?
Anyway, I come from the Open Source world and I didn't ever see a situation with a so bad user support. Hey Mr Corel, where are you? How can we get in touch with you?
RB

[fcoiffie]

I confirm also that AutoLevel and AutoContrast are broken in ASP.
I used to use it in Bibble5 and ASP has absolutely not the same behaviour (and of course, even if the same values for levels) : the effect is really too strong.

[thufor]

Indeed, both seem to be broken. As if the functionality was applied to too small image in the mipmap... If this is the issue the fix should be trivial.

[gregglee]

And if you "go from there" (I assume this means adjust the sliders after AC), what is the point of doing an extra AC step? Just adjust the sliders.

It is significantly faster. Often enough no further adjustment is needed or only one for the triangle needs to be moved after pressing AC.

Not my experience, but could well depend on camera and image composition.

[gregglee]

The question is : how can we report this bug to the dev team? By means of the Bibble forum? Or is there another way?

I reported it on Jan 17 using https://www.surveymonkey.com/s/ASPFeedback

More reports might help the ranking.

[Jeff Stephens]

More reports are not necessary. We're aware of this; it is not trivial to fix, but it is on our radar..

Cheers, Jeff

[gregglee]

More reports are not necessary. We're aware of this; it is not trivial to fix, but it is on our radar..

Cheers, Jeff

Thanks for some feedback.

Certainly it is not a trivial problem. It's a serious bug in a basic function.

But it is not obvious why "it is not trivial to fix." (I assume that means "It is difficult to fix.") It worked well in Bibble, so it would seem like the solution is known.

I am uncertain what "it is on our radar" means. Are you aiming the missiles for launch, or watching to see if it turns and goes away.

[Tadjio]

More reports are not necessary. We're aware of this; it is not trivial to fix, but it is on our radar..
Cheers, Jeff

Thanks for your input, Jeff.
It is reassuring to know that you take this issue seriously enough to post yourself
as well as the fact that you are reading and responding to our discussion