Fig. 17
From: Line and circle detection using dense one-to-one Hough transforms on greyscale images
50 best circles of the different algorithms in different conditions. Row 1: the SHT uses exhaustive one-to-many voting on Canny contour image (parameters of the Canny detector are σ=2, t 1=6, t 2=3 for column 1, σ=2, t 1=6, t 2=4 for column 2 and σ=1.5, t 1=2, t 2=1 for column 3). Row 2: the DHT V2 (two-pass) uses three scales of estimation, and dense voting weighted with Frobenius norm of the Hessian for columns 1 and 2, and uses two scales and uniform weight for column 3. Row 3: the 2–1 HT from OpenCV uses two-pass voting on Canny contour image. Row 4: the RHT uses many-to-one voting on Canny contour image. Row 5: the CACD uses one-to-one voting on Canny contour image. (For those three last methods, Canny’s parameters vary like the method of row 1)