Image smoothing and edge detection by Hermite integration

摘要

In the present paper, we propose Hermite polynomial integration method for Gaussian and Gaussian derivatives filtering in image smoothing and edge detection. Hermite integration method for one dimensional (1D) Gaussian filtering is at first introduced, and its performance is analysed. The Hermite method is then generalized to Gaussian-derivative filtering and to multi-dimensional cases. Application for edge detection with a subpixel precision is presented as well. Experimental results, which confirm the theoretical analysis, are reported.