Comparison of various centroiding algorithms to determine the centroids of circular marks in images
Rakesh Gohane, Sagar Adatrao, Mayank Kumar Mittal · 2016
Determining the centroids of circular marks is a problem that arises in many applications, for example, fluid mechanics, computer graphics, coordinate meteorology, statistics, etc. The accuracy of determining the centroids of circular marks in images is important for the overall accuracy of measurements. Therefore, the aim of the present work is to determine the centroids of these circular marks with reasonable accuracy. Artificial images were generated with the variation of image signal-to-noise ratio (SNR). First, the images were segmented using basic global thresholding method. Then different centroiding algorithms, namely, Center of Mass (CoM), Weighted Center of Mass (WCoM), Spath algorithm and Hough transform, were used, and RMS errors in centroid detection were compared. A circular mark with radius of 8 pixels (reported as commonly used value in literature) was used to study the effect of varying SNR. Results showed that CoM and WCoM techniques gave higher accuracy in centroid detection for high SNRs, whereas Hough transform performed better for low SNRs. However, Spath algorithm showed comparable errors to those of CoM and WCoM techniques for high SNRs and those of Hough transform for low SNRs. Hence, in overall sense, Spath algorithm is the most accurate algorithm for all SNRs.