**4. Experiments**

276 Fuzzy Inference System – Theory and Applications

22. IF (M is high) and ( hDV is medium) and (HP is low) THEN ("Edge" is medium). 23. IF (M is high) and ( hDV is medium) and (HP is medium) THEN ("Edge" is medium).

24. IF (M is high) and ( hDV is medium) and (HP is high) THEN ("Edge" is high). 25. IF (M is high) and ( hDV is high) and (HP is low) THEN ("Edge" is medium). 26. IF (M is high) and ( hDV is high) and (HP is medium) THEN ("Edge" is high). 27. IF (M is high) and ( hDV is high) and (HP is high) THEN ("Edge" is high).

 (a) (b) (c) (d) Fig. 6. Using different methods to detect the image with gradation of gray levels

The proposed fuzzy edge detection method is simulated using MATLAB 7.0 on different images, and its performance is compared to that of the Sobel and LoG operator. In Figure 6, (a) is the original image, (b) is the image that the Sobel operator with threshold automatically estimated from image's RMS value. (c) is the image with LoG operator to detect edges, and the threshold is computed automatically. The FIS system, as shown (d), not only detects edges much better, but also makes the output image without noise.

Form above the experiments, it can be obviously shown that no matter how different images are tested, such as from bright to dim or from natural to artificial , the FIS system proposed in this paper is much better than Sobel and LoG operator in edge detection. The only disadvantage is that FIS system is not as simple as Sobel and LoG operator.
