An Improved Method for Circle Detection

The algorithm of Hough transform for edge detection is very useful in digital image processing because of its insensitivity to the noises in the source image, so perfect results can be achieved when the dimension of parameters is less than two. However, the difficulty and complexity of calculation will be increased if the dimension of parameters is more than two. This deficiency can be solved by the Randomized Hough Transform (RHT) algorithm. By analyzing the probability of sampling, an improved algorithm based on the Randomized Hough Transform is proposed in this paper. Some proper methods are used in the step of noise filtering, so the efficiency of the program is increased. The center positioning result is at a sub-pixel level, and also, non-circular objects can be eliminated accurately by using this algorithm. All these advantages can be proved through the experiments in this paper.