SCALE-SPACE RANDOM WALKS

摘要

The Random Walks image segmentation algorithm provides a fast and effective method for supervised image segmentation. However, Random Walks does not work very well in the presence of noise or texture. Therefore, we propose an augmented version of Random Walks known as "Scale-Space Random Walks" (SSRW) that addresses these problems. Through a minor, though non-trivial, modification to the Random Walks algorithm, we show that the SSRW can produce more accurate segmentations in the presence of noise and texture then the original Random Walks can.