One Approach To Multi-scale Edge Detection

M. Kazanov (Russia)


Multi-scale segmentation, edge detection, image processing


A new method of multi-scale edge detection is presented in the article. This method uses a fine-to-coarse tracking strategy to achieve high positional accuracy of edges. High accuracy is very significant in many practical problems such as character recognition.We propose an approach that prevents the already detected sharp edges from smoothing while moving to coarse scales. The introduced algorithm allows in the same time to detect successfully ramp edges at coarse scales. Unlike other multi-scale edge detection methods that discern only edge po sitions, our approach yields a multi-scale gradient image that can be used for further analysis or in segmentation algorithms. We also present a modification of this method for noisy im ages based on watershed transformation.

Important Links:

Go Back