v4.80 autodetect brightness adjustment

Helicon Filter: Want new feature? Suggest it here

Moderator: Stas Yatsenko


v4.80 autodetect brightness adjustment

Unread postby Guest » 20.10.2007 17:43

My observations after a few minutes with the new Brightness options:

I'm having mixed thoughts about the Autodectect funtion.

The new autodetect algorithm doesn't work well for the underexposed images I'm testing. The old algorithm worked better for me (though the local contrast setting was wrong in v4.71).

Having autodetect as a button was fine as a starting point. Having it as a checkbox seems promising but would need to tuneable if it's to be more than just a starting point for each image. I don't mean corrections at absolute levels you do after autodetect to tweak an individual image, but corrections which could be applied to a whole folder. If applying to a whole folder, the corrections need to be something other than absolute levels of adjustment, more like % of effect, or weak, normal, strong.

But this starts to get complicated.

Perhaps a better approach would be to separate Autodetect into:
- auto exposure (automated Gamma level)
- auto contrast (we already have this)
These could be used separately or in combination.
As with the current auto contrast control, auto exposure could benefit from multiple levels too. I can think of two approaches... setting desired EV level (scene average), or perhaps better % Gamma correction toward 0EV scene average, assuming the original picture metered the way it did for a reason (e.g. black background); the % correction would make the auto exposure control work just like the auto contrast; 25%, 50%, 75%, 100% might work fine. I think this keeps things nice and understandable.

I'm making the further assumption that application to a whole folder is application of the % settings, not any absolute Gamma or levels or contrast settings which may have resulted from the image in view when these controls were set.

Either of these corrections would benefit from some saturation correction also built into them as they do affect apparent saturation for sizeable corrections. Such saturation correction needs to be scaled to the size of the brightness / contrast corrections. In the case of Gamma correction, I experimented with corresponding saturation adjustment and found that x points of Gamma increase required ~ -x/2 points of saturation correction to keep appearances similar (except for exposure).

One could always put the Autodetect button back and have it simply select 75% autoexposure and 25% autocontrast as a starting point (I'd guess these would be fairly reliable settings).

Here are reduced versions of the files I'm testing.
http://i14.photobucket.com/albums/a319/ ... 232_FS.jpg

slightly overexposed
http://i14.photobucket.com/albums/a319/ ... 231_FS.jpg

Posts: 17
Joined: 16.07.2006 01:38

Let me add...

Unread postby BillZM » 20.10.2007 17:57

Oops. That was me; forgot to log in.

Prior Autodetect (v4.70 and earlier) mixed in a few points of local contrast with the autocontrast setting. This was needed to knock down the increase in contrast from the levels adjustment evident with underexposed photos anyway.

The autocontrast slider itself doesn't do that; it only seems to adjust black and white levels. Under the scheme I suggested above, some similar adjustment might still be required.

User avatar
Dan Kozub
Posts: 1157
Joined: 24.03.2004 18:14

Unread postby Dan Kozub » 21.10.2007 21:11

Thanks for your comments!

I have experimented with my latest photos (http://photo-os.com/dan/thumbnails.php?album=61&page=1)

and fine tuned the sliders to work good in most case but not too aggressive. Once you lost highlights, you will never get it back. So I decided to set quit conservative values and correct only shadows. Using Gamma slider often makes sky completely overexposed.

So now Autodetect should bring you to the good starting point for further corrections without loosing info in the highlights.

I do not want Brightness tab overloaded with too many sliders. I am thinking about adding some page to the preferences where you can setup such things.

Posts: 17
Joined: 16.07.2006 01:38

Unread postby BillZM » 22.10.2007 04:08

Thanks, Dan. Are you referring to what's in v4.80? Or experimentation you've done since? If since, something we'd expect in v4.81?

I love the balance of power vs simplicity you've created in Helicon Filter and your conscious consideration of workflow. It's easy for all of us to think of things we might change, but I'm grateful that you carefully consider each comment and strive to maintain this balance.

My comments above might prove useful or they may not, but I thought share them.

User avatar
Dan Kozub
Posts: 1157
Joined: 24.03.2004 18:14

Unread postby Dan Kozub » 22.10.2007 21:16

Most probably we will prepare 4.81 version with Autodetect slider similar to Autocontrast slider.

You will have the possibility to select the "aggressiveness" of the detect algorithm then.

If you have suggestions to this function, I'd appreciate it.

So far we are using only Shadows slider and levels (black and white points). I do not like Gamma effect as it compresses highlights too much and the details in the sky are often got lost. This is not good for automatic corrections. So we use now above sliders as most universal and non-destructive tools.

Posts: 17
Joined: 16.07.2006 01:38

v4.81 Autodetect still not optimal

Unread postby BillZM » 04.11.2007 19:30


This still isn't working well on my two test files (one underexposed, one overexposed). Can I send these files to you?

As I mentioned in an earlier post here, I still think the best automated features you can do on the brightness tab are autocontrast and autoexposure.

The current autocontrast slider is perfect. Please don't change it.

Rather than autodetect, I think you need a second slider for autoexposure. Probably also allowing 0,25%,50%,75%,100%. The user could select any combination of autocontrast and autoexposure sliders. The x%/y% settings would be maintained across the whole folder if desired as %'s (not as values determined on first slide as I noticed autocontrast was doing recently).

Also I don't understand why the manual exposure correction doesn't maintain linearity between black and white levels. It should I think. You'd want linearity for the autoexposure slider too.

These are my thoughts.

How would I send you my test files?


User avatar
Dan Kozub
Posts: 1157
Joined: 24.03.2004 18:14

Unread postby Dan Kozub » 04.11.2007 22:27


Thanks, I got your files.

In my understanding the auto exposure function should not kill the highlights. This is unrecoverable damage which could not be corrected later by manual settings.

So the first "safe" operation is to clip highlights to fully use dynamic range. I usually shoot with -1...-2EV settings to be sure the sky is not clipped by the camera. And then I need to "normalize" the image in Helicon Filter, which can be done with Autocontrast function.

Other changes to "normalized" image can be only nonlinear compressing some parts of the dynamic range (usually highlights) and extending others (usually shadows). Actually it is tonemapping operations.

Autodetect function uses Gamma and Shadows tools to make dark images brighter and pull shadows out. Exposure tool is used to prevent bright areas from being clipped due to use of new Zone size slider.

Do you have better ideas for automatic exposure correction?

Posts: 17
Joined: 16.07.2006 01:38

Unread postby BillZM » 09.12.2007 20:12

Sorry for the delay. I had a lengthy reply typed out a while back and somehow lost it through connection problems.

The short answer: with no other adjustments, for underexposed photos, I find your autocontrast slider set to 25% to be most of what's needed. On my files, the result would be better if (as looking at the equalizer) response was maintained as linear between black and white points.

I'm not sure how to process overexposed photos. Underexposure is my main issue.

Return to “Helicon Filter: Suggestions and improvements”

Who is online

Users browsing this forum: No registered users and 1 guest