Tutorial 1

  1. Consider an image of dimension 100 x 100.  Compare the memory requirements to represent the data using

    1. a structured point dataset
    2. an unstructured point dataset


  2. When designing an interpolation kernel, the various issues that one must consider are:


    In the last issue above, it is important that the kernel function is so designed that the sum of all the function values evaluated at the sampled points is 1, as these function values would be used as weights in the interpolation process. Ideally, the kernel function should be continuous and smooth also.

  3. Show that the linearly interpolated attribute value, f(Q), of any point Q inside the triangle formed by 3 sampled points P0, P1, and P2 is
    f(Q) = w0f(P0) + w1f(P1) + w2f(P2)
    where
    w0 = (1-r-s)
    w1 = r
    w2 = s
    The two scalars r and s are defined as follows: r = ||P0Q1|| / ||P0P1 || ,  s = ||P0Q2|| / ||P0P2||,   where QQ1 is parallel to P0P2   and QQ2 is parallel to P0P1,  as shown in the diagram below:

    Bilinear-3pts.jpg