Multi-Scale Image Segmentation Using Watershed-Based Transform
In this paper we introduced the technique of multiscale image segmentation using watershed-based transform.We firstly use flooding watershed technique to obtain the over-segmented regions (objects),and then employ a region-based merging algorithm to construct a hierarchical network of image objects according to the scales of object.The time to stop the procedure of the region-merging depends on the average size of all object regions,and its threshold can be computed by the function of the input scale value.Finally,we convert all resulting segments to vector-format polygons by Douglas-Peucker (DP) algorithm.Experiment results show that our method is an accurate,doable way for color image segmentation.
component flooding watershed techniqu region adjacency graph technique hierarchical network of image objects
Ke LU
Department of Computer and Information Engineering,Zhejiang Water Conservancy and Hydropower College Hangzhou,310018,China
国际会议
太原
英文
309-312
2011-02-26（万方平台首次上网日期，不代表论文的发表时间）