A novel method for circular objects extraction based on region information

By using region pixels rather than edge points, a novel method is proposed for extracting circular objects in an image. A two-dimensional accumulator array with less columns is voted by region pixels instead of contour points. A set of quadratic functions and a linear function are fitted, and the coefficients of the fitted functions are used to compute the circle parameters. The quadratic polynomial function, which is determined by the circle radius and mean voting distance, is deduced by analyzing the relationship between voting values and the voting distance. A linear function, which is determined by the circle center, is deduced by analyzing the relationship between mean voting distances and the voting angle. In each column of two-dimensional array, a quadratic function is fitted and validated. The circle radius is computed based on the coefficients of the validated quadratic functions. After all columns are processed, a linear function is fitted. The circle center happens to be the coefficients of the fitted linear function. The proposed method is tested and compared using synthetic images and real-world images. Experimental results show the proposed method is fast and accurate in the presence of noise disturbance.