image to gcode

That image would probably be a lot better converted to a vector format and then run through something like a DXF to G-Code converter.
I think that Inkscape should be able to vectorise the artwork.

eslavko wrote:

The problem is that in this way for each line I got two vectors (left and right side) but I want thick line to be thicker (deeper with V cutter) and not just two outline's.

This was discussed in the context of carving text on IRC. Not that any conclusion was reached.
You would need a preprocessor that moved a variable-diameter circle so that it is always tangent to two lines.
There are commercial packaged that I think do this. There is one I see advertised around the place (seems to be Window only, but I guess might run under Wine) that allows you to download a demo: www.vectric.com/WebSite/Vectric/vcp/vcp_index.htm
I know nothing about it other than that it exists.

The image-to-gcode works better in this in mind (not perfect but...)
Is there any way to overcome that? (I use inkscape 1/2 hours total).

You might get results closer to what you want by blurring the original bitmap image to increase the range of get values and the width.

I guess that the algorithm is simple in concept but tricky in practice.

The path is the centre of a circle that is tangent to the line being followed. The Z-height is chosen such that the circle touches another line (and all touched portions of that other line are deleted from the list) or the max depth for the cutter/material is reached. Alternatively, I suppose that you could default to minimum depth for such lines.

Choosing which side of any given line is the "inside" and which is the "outside" will be a fun puzzle.

The "list" I mentioned earlier might want to be prepared by breaking the curve down into short straight lines, then there are (I think) fairly efficient algorithms for finding lines that touch circles.

You might, for example, take the start point of a segment and walk the list looking for the line segment which has the shortest perpendicular distance to that point. The midpoint of that perpendicular line is a point on the cutter path, and the length of the line is the cutting circle diameter (which sets the z height). You then solve again for the other end of the segment, and add the path between those two points to the cutter path along with the two Z heights.

I think you might also split the "found" lines at the perpendicular points and delete the intermediate segments. But that needs a fair bit of thinking about.