Recognition of Lines Detected in the Image Plane on the Basis of the Generalized Spectral–Analytical Method

摘要

In recognizing visual images, it is important to have a quantitative characteristic of similarity between lines detected in the image plane (such as closed contours, gradient discontinuities, curves, etc.). To obtain such a characteristic, we vectorize each line and find the dependence between the direction angle of its traversal vector and the traversed path length (the course function). The generalized spectral–analytical method (which consists in completely processing data in the space of the Fourier coefficients obtained by expanding the course functions in orthogonal series) provides the means for fast estimation of the similarity of lines with various locations, sizes, orientations, symmetries, etc. The effectiveness of this approach is tested in the example of Chebyshev polynomials (of a discrete argument).