A Novel Multiscale Edge Detection Approach Based on Nonsubsampled Contourlet Transform and Edge Tracking

Table 3

The and values, where the 1, 2, 3, and 4 denote top, middle top, middle bottom, and bottom noise images in Figure 13(a), respectively.

Noise images		Comparison to other methods
		Canny	LoG	Sobel	M-Sobel	Our method
1		0.1825	0.2004	0.8210	0.8406	0.9404
		0.1561	0.1806	0.7916	0.8343	0.7918
2		0.2002	0.3086	0.5582	0.5848	0.5944
		0.1540	0.2617	0.5611	0.5458	0.5099
3		0.4488	0.5028	0.5368	0.5491	0.6876
		0.3903	0.5920	0.4068	0.5037	0.7349
4		0.3402	0.6992	0.8078	0.8083	0.8686
		0.2535	0.6756	0.8151	0.8405	0.7841