engleski

A Curve Matching Algorithm for Dynamic Image Sequences

In this paper, a curve matching algorithm for the detection of moving object has been developed. The complete procedure of curve extraction, registration, tracing and matching is described. Curves are represented using the 3rd order polynomial approach. Searching window in the successive dynamic images of the observed video sequence is narrowed using the motion estimator. In each image, the curve of winner chain is selected by a novel algorithm based on calculating the highest similarity coefficient. The similarity coefficient is the compilation of multiple factors. The proposed algorithm is evaluated through tests that show high reliability in real world problems.

Curve matching; Similarity coefficient; Dynamic images; Winning chain

nije evidentirano