Jason Stredwick


Current Residence:
Bothell, WA 98011

CV / Resume
Masters Work
     Temple and Fountain
     Monte Carlo Ray Tracing
     Various Graphics
     Socket Shifting
     Mutation Rates
   Various Code
Misc. Ideas
Source code and executables are provided in the zip file.


Summary / Readme

This is exploration 1 for cse891 Advanced Computer Graphics. All four questions I chose were merged into a single situation. The four questions I chose to answer are: curve that touches all of its control points, objects that move along a curved path with independent posing, moving a polygon mesh along a path defined by a series of control points, and to deform a mesh.

The curve function I used in order to touch all of its control points was the Catmull-Rom algorithm. At least the one I used, is based on the Ferguson algorthm which takes two points and the tangent at those points. The Catmull-Rom is a simple extension of the Ferguson algorithm in that it computes the tangents for you by taking two additional control points instead of explicitly using tangents. The extra control points are supposed to be to the left and right of the primary control points. The tangents at the primary control points are computed by subtracting the point on its right from the point on its left creating a tangent line. Thus if the points are p_left, p0, p1, p_right, then the tangent at p0 is (p1 - p_left) / 2 and at p1 is (p_right - p0) / 2.

Most implementations state that it requires a minimum of four control points, however this is not true. Three is the minimum number of required control points. The reason for this is that this algorithm does not handle the end control points well. Special cases are required to handle them. The way I decided to handle the end control points is two break it into two cases: control points that loop, and those that do not loop. If the control points loop, the minimum required number of control points is three in that for the two primary points they share the same point as their right and left points. Thus if you have the points p0, p1, and p2, and computing the curve between p0 and p1 you would call the Catmull-Rom algorithm with p2, p0, p1, p2. If the control points do not loop, then I extrapolate to the left or right respectively. The following is the calculation for the left and right points should they both need to be extrapolated:

p_left = p0 + (p0 - p1)/2 and p_right is p1 + (p1 - p0)/2.

For the representation of the spheres from Phantasm(exploration requirement), I use the Catmull-Rom algorithm to generate a corkscrew path for the spheres to follow. The spheres take a start and end position where the start position denotes the local origin of the path and the end position is where it should go. Each death sphere is a sphere with a cylinder, scaled to a cone shape, coming out of it. The sphere moves along the corkscrew path and its orientation is determined by its current location and the end position. A matrix is computed to determine this orientation based on those two values.

For the moving tube portion of the exploration, I generate a simple tube using control points and the Catmull-Rom algorithm. The path loops over a set of control points and uses the Catmull-Rom algorithm to compute the actual curve. The tube is created by breaking it down into segments. Each segment is a specific distance along the path of motion behind the front of the tube. Each segment is then subdivided into rings whose centers fall on the path of motion and oriented in the direction of the tangent of the path at that location. Each circle is subdivided into a number of vertices and these vertices are connected to those in the rings following and preceeding rings to form a triangle mesh.

For the deformation portion of the exploration, I did not use a facial model but used my mesh tube from the other question. The scenario devised takes two cylinders with a distance between them that is smaller than the width of the tube. To make this scenario more visible, The collision volumes for the cylinders are larger than the actual cylinders. This allows you to view the deformation easily without looking like the tube was just going right through the cylinders. To do the deformation, I first compute a triangle's coordinates. Then I test for intersection of these new points with the cylinders based on the origin of the curve, since the control points form a circular type curve. If a new point would have intersected the cylinder, the point is adjusted to be at the location where the ray from that point from the curve origin would have intersected the cylinder. Using these adjusted points, you get the affect that the tube is deforming between the two cylinders.

* Please note that I did not complete the ant skeleton question. While the code is available for its current state, it must be part of the Visual Studio project or it will crash due to some internal MFC error.