Abstract
This paper proposes a robust and fast Hough transform by which a set of lines can be detected by the use of high-speed M-estimator. This proposed method is one of the matching methods that is similar to the least squares (LS) method, however, the proposed method does not use the conventional parabolic function for the error function, and instead, does use the 2nd order B-spline basis function so that the following extreme advantages could be provided : The ill effect caused by the outlier will be drastically decreased in the proposed method, and in addition, the estimated line becomes just the same as those given by the LS fitting when no outlier exists. Furthermore it was possible to reduce the computation cost by introducing the successive usages of the rectangular filter.
