how is this ?
Good luck with that 
I guess that looks easy, but I’m not sure it looks useful.
AHHH, No! Not the broken topology and huge poly count. This method is asking for trouble. If you need fine surface detail try using a normal map instead. You could always try blenders new dynamic retopology in sculpt mode but don’t forget to retopo :eyebrowlift2: In the cross model you only need 4 vertices for each of the long extrusions, can subsurface or bevel if it still needs more detail after that.
It looks like a complicated way to simulate voxel modeling to me. If voxel modeling is what you want, then I’d recommend using a voxel modeler (if those exist, I haven’t really looked into that) or using the remesh modifier on a standard poly model.
I don’t see how it could be useful, i think it just adds unecessary verticies.
Subdivide a plane 1000 times, add a background image and select all vertices inside the image?
That’s 2D projection, you should do it also with the front and top view of the dog.
You should have an internally subdivided cube, then circle-deselct the outer verticies. But then you would have unneeded edge and faces inside the cube, and also this method doesn’t work with concave superficies
I’m still trying to figure out A) what one would ever use this for, and B) what would make it better than any other method for that goal.
Well there is a few reasons why you would use this, but not for serious modeling. For example displacement maps can be a good starting base for terrain, an example being the CryEngine. I sometimes use procedural displacement for varying a mesh for some types of organic models such as trees and rocks. Not to mention its great for some types of animation, However that is a little bit different since it does not involve UV’s or 2D raster.
It can be better because its extremely fast in some cases. However for the image posted on this thread I still stand by my original post in this being an all around bad idea.
You are right.by the way any way its can bee useful