Abstract:

We propose a multiscale wide line detector based on isotropic nonlinear filtering to overcome the shortcomings of the basic wide line detector. The basic detector extracts wide lines using a single-scale circular mask, as a result, it can only detect lines thinner than the mask radius and has poor detection performance at intersections of two wide lines. Different from that, the multiscale wide line detector computes the response with several circular masks of different radii, and chooses the maximum response among the normalized detection results on different scales. Finally, poststeps such as thresholding and filtering are designed to remove interference and obtain the final results. Simulated and real images are adopted for performance test of the multiscale wide line detector. Experimental results demonstrate that it outperforms the basic wide line detector and can effectively detect wide lines in images.

Key words: wide line, detector, circular mask, multi-scale