S. Khairunniza-Bejo, H. Ibrahim, and A.R.M. Shariff (Malaysia)
Local similarity measure, change detection, imageclassification and image thresholding.
In this paper, the performance of point similarity measure in multitemporal data of satellite images will be presented. Mutual information will be used as a measure of similarity. Therefore, every pixel will be assigned to its own value of mutual information. These pixels are then grouped into four different groups of similarity. Sequence of local mutual information images will be produced. Image thresholding is then applied to the local mutual information image taken from the longest duration of gap. Overlapping part between the thresholded image and other local mutual information images are then analyzed. Combination of the other local mutual information images at the overlapping part is used to produce another version of thresholded image. This image is produced by referring to the change and unchanged conditions. Both thresholded images are then compared. From the experiment, it has been shown that pixel with high value of local mutual information give more consistent result.
Important Links:
Go Back
