Remote sensing image segmentation based on Dynamic Statistical Region Merging

Using the Statistical Region Merging (SRM) for remote sensing image segmentation, we found the result is unsatisfactory. To improve segmentation accuracy and the correctness, this paper proposed a Dynamic Statistical Region Merging (DSRM). It tries to let the most similar regions to be tested first. At first, it redefines the dissimilarity based-on regions. Then, it dynamically updates the dissimilarity and adjusts the test order during the procedure of merging. Experiments demonstrate the accuracy of the DSRM is higher than the SRM and its computational complexity is approximately linear. In addition, we extend the DSRM to multi-band remote sensing image and use it for multi-scale segmentation.