Abstract: The traditional edge detection methods have high sensitivities to noise, a novel color image edge detection algorithm based on mathematical morphology is proposed. Gray-scale morphology is extended to color image by sorting each pixel as a vector in RGB space. And then by analyzing the characteristics of the noise (mainly salt and pepper noise) contaminating image, color image morphological basic operators are improved. Improved operators are insensitive to noise and can directly achieve edge detection. Experimental results show that this algorithm can more effectively suppress noise on the impact of edge detection and better preserve edge image details than traditional methods.

Key words: computer application, edge detection, mathematical morphology, color image