1: \begin{abstract}
2: A versatile method is described for the practical computation of
3: the exact
4: discrete Fourier transforms (DFT), both the direct and the inverse ones,
5: of a continuous function $g$ given by its values $g_{j}$ at the points of
6: a uniform grid ${\sf F}_{N}$ generated by conjugacy classes
7: of elements of finite adjoint order
8: $N$ in the fundamental region $F$ of compact semisimple Lie
9: groups. The present implementation of the method is for the groups
10: $SU(2)$, when ${\sf F}$ is reduced to a one-dimensional segment,
11: and for $SU(2)\times SU(2) \cdots \times SU(2)$ in multidimensional
12: cases. This simplest case turns out to be a version of the discrete cosine
13: transform (DCT). Implementations, abbreviated as DGT for
14: {\it Discrete Group Transform}, based on simple Lie groups of higher
15: ranks, are considered separately.
16:
17: DCT is often considered to be simply a specific type of the standard DFT.
18: Here we show that the DCT is very
19: different from the standard DFT when the properties of
20: the {\it continuous extensions} of the two inverse discrete transforms
21: from the discrete $t_j\in F_{N}$ to all points $t \in {\sf F}$
22: are studied. The following properties of the continuous extension of
23: DCT (called CEDCT) are proven and exemplified.
24:
25: \noindent
26: Like the standard DFT, the DCT also returns the exact values of $\{g_j\}$
27: on the $N+1$
28: points of the grid. However, unlike the continuous extension of the standard
29: DFT,
30:
31: \noindent
32: (a) the CEDCT function $f_N(t)$ closely approximates $g(t)$
33: {\it between} the points of the grid as well.
34:
35: \noindent
36: (b) For increasing $N$, the derivative of $f_N(t)$
37: converges to the derivative of $g(t)$.
38:
39: \noindent
40: (c) For CEDCT the principle of locality is valid.
41:
42: \noindent
43: Finally we use the continuous extension of the 2-dimensional DCT,
44: $SU(2)\times SU(2)$, to illustrate its potential for
45: interpolation as well as for the data compression of 2D images.
46:
47: \end{abstract}
48: