摘要
Based on the theory of data smooth, a new technique for edge detection is presented. It combines the 2-dimensional spline function consisting of the tensor products of 1-dimen-sional cubic B-spline with the Laplacian operator and presents a ▽2S operator. Following the image convolution, edge points are detected by the zero-crossing of the output. This operator proves better than the ▽2G operator presented by Marr-Hildreth in image smoothing and edge detection, and it requires less computation work. It is shown that ▽2G can be approximated by a higher-order ▽2S. Some experimental results are also given.
Based on the theory of data smooth, a new technique for edge detection is presented. It combines the 2-dimensional spline function consisting of the tensor products of 1-dimen-sional cubic B-spline with the Laplacian operator and presents a ▽2S operator. Following the image convolution, edge points are detected by the zero-crossing of the output. This operator proves better than the ▽2G operator presented by Marr-Hildreth in image smoothing and edge detection, and it requires less computation work. It is shown that ▽2G can be approximated by a higher-order ▽2S. Some experimental results are also given.
