An Improved Random Walker using Spatial Feature for Image Segmentation

The random walker algorithm for image segmentation is an effective method. However, the traditional random walker algorithm (RW) is only determined by the adjacent node (pixels) intensity information in the feature space, and the algorithm does not take the spatial feature information of nodes (pixels) into consideration, which makes the segmentation results discrete in the spatial distribution. In this work, an improved random walker algorithm (SRW) has been proposed to improve the efficiency and accuracy of the extraction with complicate background. Firstly, spatial feature information has been employed to combine with the intensity information to measure weights between adjacent nodes (pixels). The freedom parameters have been then adjusted for the two features above to obtain the scale. Finally, the experimental result shows that the improved random walker algorithm (SRW) is effective in extracting objects contour with complicate background.