What exactly are halos?

[bbodine9]

I guess I do not know what I am looking for but what exactly are halos and what is their primary cause? To get rid of them what is the proper procedure? Are all artifacts halos? Thanks!

[ksinkel]

There are lots of different kinds of possible artifacts, halos among them. Halos are just what you would think they are -- a light area around a dark area produced by excess manipulation. In general, if an artifact is not evident, then the manipulation is not in excess. If it is, you generally need to back-off. Sharpening is a good example of a manipulation which is lmited by the artifacts it can produce.

Kiril

[Dieter Mayr]

Halos likely appear when exeeding with sharpening.
Left is the original image, right is with 2 times unsharp mask, amount 100 % and radius 10:

Halo-Demo.JPG (22.41 KiB) Viewed 8039 times

I made it extreme in this example of course, to show the effect, depending on the background,
i.e. a clear blue sky, a much smaller effect is visible.
Halos can also happen due to spherical aberration in lenses, but recent lenses are normally
very well corrected.

[den]

Not so obivous, but 'halos' can also be 'dark' on 'light' backgrounds:

halo_1.jpg (48.6 KiB) Viewed 7868 times

As previously mentioned, halos can be produce by overly aggressive sharpening or as illustrated directly above, overly aggressive 'Local Contrast Enhancement'... and again, the solution is to back off on the amount of sharpening or LCE to a more photo-realistic appearance.

Halos can also be produced by making aggressive tone changes when using masks that have symmetrical gradients from a 50% tone between mask white and mask black... a solution here is too again back off on the amount of change or if the original mask is still available... make a 'mask of the mask' and use the new mask to decrease contrast/saturation in the halo image areas...

...or...

Use an asymmetrical mask where the gradients from the 50% tone between mask white and mask black differ. A discussion/examples appear here:
http://www.ncplus.net/~birchbay/tutoria ... /index.htm

[tomczak]

As Den mentioned, it's easy to introduce halos when sharpening e.g. using Unsharp Mask. Such halos are symmetric about the 'edges' - the dark part of the edge is made even darker and the light part of the edge is made lighter. This, in principle, should work to increase our perception of sharpness by increasing the acuity of the edges and details. The halos are not equally objectionable though: the white halos pronounced edges (e.g. tree trunks against the light sky) are probably the most annoying. One technique that can help to reduce this effect was suggested by Bob Walker some time ago, and it works fine in many situations.

- unsharp mask
- Composite the original image with the sharpened one as an overlay, Operation=Darken

Another way of avoiding halos is using the new Bilateral Sharpen. You can set it up so that the major edges are protected from producing halos (i.e. excluded from mean-image blurring) while lesser edges/details (usually inside a some areas defined by major edges) are sharpened (or their local contrast is enhanced).

[MikeG]

Maciej,

On my machine, sharpening a 12 Mega pixel jpeg using Unsharp Mask takes approx 5 seconds.
Using bilateral sharpen takes approx 6 times as long - over 30 seconds.

Have you a technique using Bilateral Sharpen that significantly reduces the processing time while maintaining most of the benefit? Or perhaps you're more patient than I am...

Mike.

[tomczak]

As it is now, the Mean Image is calculated for all pixels, regardless of what's visible in preview. During beta testing, I recall Jonathan mentioning that it may be possible in the future to calculate mean image for the preview only, do adjustments, and leave the final calculations till the end. I believe that bilateral sharpen is computationally intensive no matter what you do, but moving the waiting till the end would make adjustments much easier - and this is such a neat transformation. I think the calculations in USM or AS are optimized this way - small 1:1 preview allows for v. fast adjustments.

What I do now is simulate just that: I crop some small but significant training portion of the image, do the Blur Radius and Blur Threshold adjustments on this crop to my satisfaction (imagining what will it do to the rest of the image), then switch to the actual image, click Preview (or OK) and wait. Adjusting Sharpen Factor and Sharpen Threshold doesn't recalculate the mean image so they are fast on any image size and can be adjusted at any time.

[MikeG]

Thanks. I'll try that, looks like a good approach.

Mike.

[MikeG]

Inspired by Maciej's tip on how to experiment with the blur settings in Bilateral Sharpen with minimum time penaly, I did so on a number of images and then used Bilateral Sharpen as part of my project 'Don't buy a macro lens - you don't really need one".

Following are the original photo. Taken with a Panasonic G1. 14-45mm (28-90mm 35mm equiv) lens at 45mm. Approx 300 mm (12") camera to feather. Feather lenght approx 25mm (1"). Forced built in flash. RAW.

And a crop from the developed RAW. The crop is 1/4 linear. 4000 pixels wide to 1000 then resized back to 4000.
Finally bilateral sharpening.

I think I've saved some money!
For anybody wondering, the feather originally belonged to a Rainbow Lorikeet.
Mike.