I'm a web dev. I guess I'm a bit (a lot) behind the times still using CS4 and CS5. I primarily work in Photoshop and I'm trying to get better with Illustrator. One -very- common task I would -love- to be able to do efficiently in AI is something like this:
I get a bitmap image, which to the naked eye is pretty much just straight lines. I want to import it into AI and then start tweaking. If I hand draw something with a sharpie and then scan it, it works fine. But very few bitmaps, regardless of how simple, Live Trace in a good way.
I've bought a couple of pricey AI books and they -always- use some really complicated drawing to show how wonderful Live Trace is. And that's the irony: the more complex the bitmap, the -better- it seems to do. It's SIMPLE stuff---like -I- use everyday for web icons and buttons that it does so poorly.
Any bitmap that has some sort of 'shaded' edge... OR is not a straight 90 or 0? fawgeddaboudit.
Always lots of swirls and curves. And then I get frustrated trying to 'simplify' paths.
SO: Are there any 'ninja' secrets to doing this? I've read a couple of posts which basically say 'man up' and learn to hand draw better, but it just seems like AI is so sophisticated that it -should- do this properly and this =must= be user error.
I'm not above girlie-man begging.
It's rarely 'simple'. Frankly, after 5 years, I am almost totally passive aggressive with AI.
I came over from Corel which had totally drop simple tools for doing what I need to do... ie. drawing nicely aligned geometric type objects. For example, it took 10 seconds to do a box or triangles or anything with -symmetrical- geometry. You'd hold down the shift key and auto-magically get 30, 45, 90 perfect arrows and perspective correct rectangles, conic sections, etc. It was just 'smarter' about stuff like that. AI feels like you -really- gotta get inside it and -then- you can do anything. But for what -I- do... where I just want to fire it up for 10 minutes do something quick and pedestrian and then get back to my -real- job? Frustrating.
So what I always hope against hope for are shortcuts to doing those sort of 'dummy' tasks... and if Live Trace worked as I hope, it would be a life saver for me.
I find myself often using PS, instead of AI, even when AI is the better tool, because, frankly, I'm just so much more comfy with it.
Maybe it's a little harder than average but to -me-, AI should have some sort of 'edge detection' to convert this to about two dozen straight 'lines'.
But what it -does- is either turn it into a mass of paisley swirls -or- if I use the 'photo' setting, the trace looks mostly like the original, BUT it's created almost 1 path for every frickin' pixel.
What I end up doing---if I really need it---is to trace over it with the mouse, but I confess I'm a LOUSY 'pen drawer' so it often goes awry. Most often? I make a print-out, trace it on paper with a sharpie, then re-scan it into AI. And THEN it will 'Live Trace' OK... seems like a total Rube Goldberg but AI seems to -need- that sharpie.
So... in lieu of going to pen boot camp, I was -hoping- that this was just operator error.
This is no sufficient basis for an autotrace. If there are only 5 pixels in a row, how on earth should the autotrace module know that these pixels form a straight line? They could instead mean anything. If this image were only a little smaller, even a human would have to guess what's on it.
How does my eyeball do it?
Experience. You know the object and therefore you recognize it. Apart from that: http://en.wikipedia.org/wiki/Gestalt_psychology
It has taken mankind quite a long time to have that. Longer than Illustrators development cycles ...
As Monica said, it is a hard case because the stripes on the barricade are not very long.
But there are some things you can do to help.
First, for low resolution images the "noise" parameter will be the one you want to play with
the most. It will affect the results more than anything else, and you will probably need to
set it lower than the default value. For example I set it to 3 to get the results below.
Second, check the "snap to lines" box. It will do what it says.
Third, move the "corners" slider to the right, I put it at 100%.
Here is what I got. Not perfect, but did get some things right:
Thanks. And thanks double plus for taking the time to try it. That's actually helpful... I dunno if it's worth $600 to 'upgrade' but it's -way- better than CS5. I admit the example I provided is a bit tougher than the typical 'icon', but from your example I think it would work just fine with a lot of the tasks I need.
AFA the other replies? Since you took the time to slap me down, -my- suggestion, if you -really- want to be helpful? 3 choices:
1) Here's how ya do it. 2) Software doesn't do it. 3) RTM: page #xxx
Telling people 'you shouldn't want that feature' or 'IMPOSSIBLE!' or 'learn to draw better!' are just not helpful... they may be how you -feel-... but that's not what I -asked-.