I was inspired by Jimmy Gunawan's blogs of monster generation, here is the first one [INSPIRATION / Pixar Monster Factory Part One](http://blendersushi.blogspot.com.au/2013/06/inspiration-pixar-monster-factory-part.html), and by search the Skin Modifier of Blender, I found the theory of Skin Modifier:
of 3D Articulated Shapes](http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.357.7134&rep=rep1&type=pdf)(Authors: Zhongping Ji, Ligang Liu, Yigang Wang). I started to think of monster model generation for game development from years ago, thanks for this paper, Dust3D is achievable now.
From my initial thought, Dust3D should be a tool like [Makehuman](http://www.makehuman.org), with more versatile features, not only can make human, but also be able to **generate monsters automatically**.
Let's start with [drawing a sphere without gluSphere]( http://stackoverflow.com/questions/7687148/drawing-sphere-in-opengl-without-using-glusphere), because I want implement the same spheres which presented in the [B-Mesh paper](http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.357.7134&rep=rep1&type=pdf).
Almost all 3D editor have a infinite grid ground, I just made a finite one, in the future, I should expand the grid outside of the screen to make it infinite.
Now, for just beginning, I think it's a not bad start.
There is a implementation of [B-Mesh algorithm in C++](https://github.com/evanw/cs224final) language, but I want the pure C version, so I start to implement my own version. I read both paper and this implementation, it gave me very helpful understanding of this algorithm.
Draw shape ball is easy, no need to rotate, I just need scale it along the ball's radius.
Draw the cylinder which connects two shape balls is more difficult, I need do some math to rotate it. [Here](http://www.thjsmith.com/40/cylinder-between-two-points-opengl-c) described it.
Camera rotate/zoom implemented, [here](http://www.songho.ca/opengl/gl_transform.html) is a good article which explained the theory of OpenGL Transformation. Most important is that the demo app is very beautiful.
When I am implementing the B-Mesh algorithm, I am also think in the future, how to create a library of bunch of initial base models. There is a paper [the Skeleton of a Closed 3D Shape](http://www1.idc.ac.il/icgf/GraphicsSeminar2006/DCGskeleton06.pdf) described how to generate skeleton from mesh, this is the reverse progress of what I am doing, I think it can resolve the problem of insufficient initial base models, I can generate from tons of existed models.
After finish the rotation at the two connected bones, I need implement 3D Convex Hull algorithm at the joint ball, there are so many methods to get the convex hull, I found the [Gift wrapping](http://dccg.upc.edu/people/vera/wp-content/uploads/2014/11/GA2014-ConvexHulls3D-Roger-Hernando.pdf) is the most strait-forward one, though is not the most efficient one.
There is a good implementation of [Gift Wrapping algorithm written in lua](https://github.com/danielhst/3d-Hull-gift-wrap/blob/master/giftWrap.lua) language. When I was implementing, I first translated this version to C language, but maybe there are some bugs on my own translation, there are lots of repeated faces, so changed a little, the code now is not just the translation of the original lua version.
There is a implementation of Catmull-Clark Subdivision algorithm on [rosettacode](https://rosettacode.org/wiki/Catmull%E2%80%93Clark_subdivision_surface/C), the code is very simple and beautiful, just lack of some memory alloc fail check. I translated the algorithm to use my own array implementation.
