K. Kpalma and J. Ronsin (France)
pattern analysis and recognition, multi-scale smoothing, curvature zero-crossing, intersection point map
In this paper, we present a new method and its preliminary results in the context of pattern analysis and recognition. Based on the curve analysis, it deals with the contour of planar objects like the CSS (Curvature Scale Space) method. This latter method uses the 2nd order derivative of a contour to determine its zero-crossing points to characterise the pattern. Like the CSS method, our method uses a Gaussian kernel to progressively smooth the curve relatively to the varying bandwidth. Instead of characterising the curve with its curvature involving 2nd order derivatives, this new method uses the intersection points between the smoothed curve and the original. As the standard deviation of the Gaussian kernel increases, the number of the intersection points decreases. By analysing these remaining points one can define features for pattern characterisation. Since this method deals only with curve smoothing, it needs only the convolution operation in the smoothing process. This way, one can reasonably believe that this method is faster than the CSS one with equivalent performances.
Important Links:
Go Back
