Why such poor results with Background Erase?

[Contours]

For some reason, I thought background erase had boundary detection logic that made it easy to isolate images. But my experience over a dozen uses has left me doubting the processing and/or the methods I use.

I'm following exactly the steps show in this article:
http://support.xara.com/index.php?/K...ckground-erase

But my resulting images never have the sharp boundaries it implies I can get.

Admittedly, the range of color (mostly grays) in the image I'm currently working with is low. Nonetheless, there is a clear edge to the image. And I've had artifacts show up even on images with dramatically darker backgrounds. Up to now, I've resorted to tediously outlining an image to get a decent result.

Is there a setting or method I should try?

Thanks in advance for any suggestions.

[angelize]

Here is a video that may help.

[Contours]

Here is a video that may help.

Actually, I had watched that very video before I made my post. It does give a good result, but the page I referenced certainly gave me the impression that a few "identifying swaths" were all that's needed ... and the software would do the rest.

Rather than try "detailed swaths", I'd much rather outline the object and be certain of what I'll get.

I was just expecting more from Xara.

Thanks for responding.

[digitalphaser]

I do not know about the software that can fulfill wishes. Xara removes background is not worse than the other software.
White background can be easily removed with "select colors" tool too.

I personally prefer the shape tool. It takes not much longer than the masking. But the clean result is guaranteed.

[Luke.Hart]

A lot of Contours problems are caused by using a JPEG original (the blocky artifacts are typical of JPEG), but this can be got around by erasing close into the foreground. For example I was able to produce a pass-able background erase of a simular image in a few minutes.

[angelize]

The image I used in my video was a jpeg but it was a fairly high resolution image.

[Luke.Hart]

Also, Angelise, you got close into the edges exactly as I described!

[Xhris]

There's definitely a knack to getting good results with this tool that experience can hone. I still feel that this tool is too trial-and-error though with its current implementation. Having to undo, retry, and hope you get a good result. It would be more accessible I feel if it you could refine the result in a live way before committing to a final result. Photoshop's refine edge functions pick out hair strands better I feel too, especially as you can alter saturation etc during the process to give a better looking result.

[Gamerprinter]

I don't know, I've only used it a couple times, both with excellent results, though the originals had 300 ppi resolution, perhaps that was the difference. I seldom use photos as-is in my cartography work, more often using photographs as vector image fills - all my shapes used in my work are vector objects. So I have little use for the tool myself, though I indeed see its practical application.

[Contours]

Thanks for all the responses. At least I know that I’m not just missing something.

Regarding magic software, we’ve had features like red eye reduction for years ... and I’ve seen some pretty amazing effects: object remove ... blending ... stitching ... content aware scaling. Even snap to objects surprises me sometimes about what it can “see”. So accurately isolating objects in an outline fashion certainly seemed feasible.

But given the realities of the current Xara feature, I’d like to see it changed it as follows:

• use the current brush method to loosely mark the object and the background
• have the software put up a dashed outline of the object it thinks it sees
• let you adjust the outline in any way that's needed, using the same tools as you would use if you had created the outline

This would let the software accomplish what it can ... and the user easily take care of the more difficult instances. And as the software improved in each new version, the user would do less and less. Plus, it’s a merging of the editing methods.

In the meantime, I’ll go with the mask + outline method for most of my uses.

Thanks, again.