itch.io

Alright Bill, I figured out the issue: there's too much image noise - with all the white spots everywhere, like what you'd see on a Xerox.  I get the same problem with the image you posted so you were right about it being the images, or at least this one (I'm assuming you have more that are also scanned Xeroxes?) PixelCNC is designed to handle both black/white binary images as well as depthmap/heightmap images, and the noise is being interpreted as details in the image, which in combination with the actual details is proving too much for default triangle/vertex buffers PixelCNC allocates at startup. As long as the noise is there you'll end up with a mesh that's riddled with little islands/bumps.

I'm going to tweak some internal settings to reduce the triangle-count a bit, which sacrifices some detail, but the mesh that generates at startup is just used for a preview of the form the image would produce if interpreted as a depthmap. I'm also considering adding some code to just re-allocate the triangle buffer if it reaches its extent - the end result here will be that a computer that doesn't have enough memory will resort to using the harddrive as a backup, but once the triangles are uploaded to the GPU for rendering it shouldn't matter anyway. At the very least it should give the user some kind of error message instead of hanging!

Here's your same image, after being cleaned up a bit. I could do a cleanup function in PixelCNC that does this automatically, but it would require the user to fiddle around with some parameters as each image is different, and some images are not black/white binary images (i.e. relief depthmaps). The user would have to fine-tune some parameters and use their best judgement. This would be the beginning of PixelCNC having a fully functional imaging component as well, the goal of which being that users can quickly design up basic images right in PixelCNC, plotting shapes and lines, gradients, text, etc..

But v1.10a simply is unable to handle the image as-is. After increasing buffer sizes this is what your original image yields:

This can still be used to do some outlining, and some V-carving:

I increased the triangle and render geometry vertex count allowed while also reducing the complexity threshold of the mesh to better accommodate larger and more complex images. I also added code to create a messagebox for end-users to see in the event of a problem, so that it's not tucked away in the log file. This way they can quickly report what happened in their initial report for a quicker turnaround.

I'll have v1.11a up ASAP.