I did a little write up on my blog for how I am creating reliefs. I will do a video of one in the future but for now, I just put it in words (mostly so I could remember 
). The page is Jobbos Blog
1 Like
This is indeed, the “holy grail” for those of us who are interested in relief carving. I’ve been going down the exact same path as Ian and Kelly for the last few weeks since I’ve gotten the MPCNC operating. I’ve evaluated various software, trying to decide whether or not I needed (or could use effectively) things like PhotoVCarve and Meshcam. (Aspire is out of the question) And, like you all, I’ve searched for some evidence that ArtCam might somehow still be out there.
So far, I’ve drawn only one strong conclusion, and that is that it is indeed a very difficult problem to automate. Because the depth information is lost completely, it appears to me that there are two factors that come into play for our human brains as we process these images: Lighting and color.
In principle, lighting might be manageable, especially if you can convert to a line drawing such as the Popeye demo guy did in the video referenced in Ian’s blog. I’ve seen several things that are sort of like that, and for a sort of “cartoon” image, his method looks promising. The attached example of the chess knight is a possibility. We know it’s color is uniform, so any apparent variation is due to lighting. If we had a line drawing, we could probably succeed. For more complex images, I’m not convinced that it will be reasonable.
The effect of color is another story. Simply converting to grayscale is not sufficient if the image is “contrasty” to start with. Consider the attached image of one of our dogs, Isis. Converting to greyscale gives you an image that is straightforward to use for a lithophane, for example, because the thicker area that is supposed to be black does indeed transmit less light, so it looks darker. However, consider what happens when you try to take that same greyscale image and carve it… We know that the black and white sections of her face are nearly in the same plane, but the computer has no idea that this is true. Thus, you will get deep holes where her nose and eyes are, as well as the tips of her ears, while the front of her face will be high. The result is absurd.
This example is extreme of course, but when I was working with it, I realized the complexity of the problem–color (i.e. tone level) complicates the process as much as lighting does.
It almost seems to me that converting the photo to a line drawing in some way and then using the grayscale gradient might be the most promising, but I’ve not tried that yet. I wonder if all these artists around the world are not doing something like that.
Finally, for anyone who has not tried it yet, you should be aware that you can do a reasonable job of the actual carving of these on-line “x-ray” looking images using ESTlcam. The process is not quite as simple and flexible as it is using MeshCAM, but if you pull one of those images into something that can generate an STL file, (such as a slicer like CURA) you can then pull that STL into ESTLcam and set up the carving. This is what made me decide that for now at least, I don’t really need MeshCAM. At least until the artwork problem is solved…
Finally getting to where I will be able to do the reliefs. Not sure if anyone has figured this out but I did come across a site that has tutorials on this:
cnc4free org
One program specifically seems promising as far as an easy approach to the height maps:
crazybump com
Just seeing this, but you can make heightmaps in blender from 3d models. you cant turn a regular image into a height map, but you can turn pretty much any STL or 3d model into one
I haven’t tried that in Blender. I will give it a go.