![The Algorithm Fuzzy c-means (FCM) is a method of clustering which allows one piece of data to belong to two or more clusters. This method is frequently used in pattern recognition. It is based on minimization of the objective function ! 1.Initialize U=[u ij ] matrix, U (0)1 2. At k-step: calculate the centers vectors C (k) =[c j ] with U (k)](https://image.slidesharecdn.com/fuzzyc-meansclusteringforimagesegmentation-100511141621-phpapp02/85/Fuzzy-c-means-clustering-for-image-segmentation-3-320.jpg)
Fuzzy c-means clustering for image segmentation
- 1. Fuzzy C-Means Clustering For Image Segmentation Computer Vision And Image Processing Course Project -Submitted By, Dharmesh Patel 961 Nikunj Gamit 954
- 2. Introduction There are several methods for segmenting gray-level images 1. Based on discontinuity 2. Based on similarity First approach use the discontinuities between gray level regions to detect isolated points, edges and contours. second approach include clustering, thresholding, region growing, region splitting and merging.
- 3. The Algorithm Fuzzy c-means (FCM) is a method of clustering which allows one piece of data to belong to two or more clusters. This method is frequently used in pattern recognition. It is based on minimization of the objective function ! 1.Initialize U=[u ij ] matrix, U (0)1 2. At k-step: calculate the centers vectors C (k) =[c j ] with U (k)
- 4. The Algorithm (Contd…) 3. Update U (k) , U (k+1) 4. If || U (k+1) – U (k) || < £ then STOP; otherwise return to step 2. Where, U = Membership Matrix £ = Termination Criteria C = Centroids X = Pixel Intensity
- 5. Membership Function K-Means C-Means
- 6. Experimental Results Source Images Output Images C = 2 C = 3 C = 4 C = 5 C - No. Of clusters
- 7. Limitations Standard Fuzzy C-Mean is not suitable for the lip and skin region The resulting regions are not spatially continuous, due to the fact that only gray level uniformity is checked.
- 8. Enhancement We can use an enhancement of fuzzy c-means, to ensure spatially continuous regions after segmentation. We can use the geometric properties of the pixels in different sized neighborhoods (typically 3x3). We can also use information about the spatial position of the pixels and not only their gray level values.