A central distance method for invariant recognition of digital figures
Teruyuki Kaneko, Tetsuo Sagara, Takashi Takeda, Ryuzo Takiyama · 2003
The authors have previously (1998, 1999) developed a description method for the recognition of non-singly connected figures using a "self-distance function" which allows for shift, scaling and rotational variation. The self-distance function is computed for all combinations of the points which make up the figures, so that the self-distance function can be calculated in O(n/sup 2/) operations via an n-point figure, which is time-consuming. This paper proposes a "central distance function" to reduce the computation time. In this improved version, we measure the distances between the "centre of gravity" of the digital figure and the points making up the digital figure, so that the central distance function can be calculated in O(n) operations via an n-point digital figure. Experimental results show the usefulness of the proposed method.