Quote Originally Posted by BeRo View Post
The common image vectorization algorithms are working without triangulation, like my unreleased but for-usage-ready image vectorization algorithm, which I've implemented some years ago. They (at least my algorithm) are working per color image quantization and then greedy from the outside inwards colorwise image contour path tracing together with incremental color plane flipping toward into the inner contour paths.
Wel I gues that then I haven't tried to use the most common approach when I was trying to implement image vectorization myself some ten years ago.
In the approach I used I first tried to convert the raster image into a scetch like image (just lines representing objects shapes) and then I was applying a triangualization algorithm to that scetch image.
It worked half of the time which was a huge sucses at that time for me. But I never figured out why it doesen't always works since I didn't know wheter the cause for that might be either in my "scetchink" code or the triangualization code which wasn't mine.
Anywhay after some time I lost interest in that firstly becouse it didn't always work and I had no diea why. Frankly even now I don't know how I managed to make it work even those half times.
Another reason why I lost interest in that is becouse it was extreemly slow. Why it was slow? Becouse at that time I was still working with Canvas.Pixel-s. And what is worse I was working with Images instead of bitmaps which made everything even worse.