An improved Prewitt algorithm for edge detection based on noised image

In this paper, an improved Prewitt algorithm for edge detection is proposed for the reason that the traditional Prewitt edge detection algorithm is sensitive to the noise. The traditional Prewitt edge detection operator only has two templates with horizontal and vertical directions. While the edge is in a plurality of directions, so operator with eight templates of different directions is put forward and it can detect more edges. In order to improve the capability of resisting noise, this paper put forward three improvements. First of all, the mean value rather than the maximum value of the gradient magnitude of the eight directions is used as the final gradient magnitude. Secondly, OTSU automatic threshold is used to set the gradient magnitude threshold. Again, an 8-neighborhood template is proposed to remove the isolated single pixel noise. The experimental results show that the improved algorithm improves the anti noise performance greatly, and detects the edges of the random noised image effectively.

Перевод пока недоступен