International Conference on Advances in Emerging Technology |
Foundation of Computer Science USA |
ICAET2017 - Number 1 |
July 2018 |
Authors: Nini, Sonika Jindal |
127e7d6e-49b5-47cf-a4ce-36e719e18843 |
Nini, Sonika Jindal . Image Segmentation using Improved JSEG with Fuzzy Weighted Moving K-Means. International Conference on Advances in Emerging Technology. ICAET2017, 1 (July 2018), 31-35.
Image segmentation is an important step towards good image analysis. In this paper, a new method of image segmentation using fuzzy weighted moving k-means and JSEG images is proposed. Although JSEG provides an efficient approach towards image segmentation but it suffers from the problem of over segmentation. In order to overcome this problem we have used fuzzy weighted moving k-means to quantize the image. In JSEG, first color quantization is done. Each color is given a label and pixels in the image are replaced by these labels. Thus class-maps are formed. Then J-images are obtained from those class maps by applying this to local windows. The technique has been applied with both moving k-means and fuzzy weighted moving k-means. In the end, the results of both these clustering methods are compared. It is seen that weighted moving k-means segment the image better than moving k-means in terms of time taken and visible improvement that is seen for over segmentation reduction.