Find in Library

This paper proposes multiscale edge extraction using the fuzzy rules. Its cardinal idea is that edges can be gained usually via extracting large-amplitude parts of signal changes in neighboring pixels. However, it is very difficult to get a threshold for distinguishing the large signal differences from small ones, because image characteristics are ambiguous and the threshold should be varied according to some local characteristics of the image. The proposed technique can perform a proper control of the threshold for separating the small-scale output relying on the value of the large-scale output. Here, the fuzzy rules are used to consider the unstableness and uncertainty of the signals. Multiscale edge extracting can obtain multiscale edges which contain more detail information in comparison with monoscale edges. Besides, edge tracing can be performed to obtain more precise edges by dividing the amplitude of the output into three components with two thresholds and making a decision with the fuzzy rules. Computer simulations verify the effective performance of the proposed edge extracting technique.