IMAGE SEGMENTATION BASED ON GRAPH RESULTING FROM COLOR SPACE CLUSTERING WITH MULTI-FIELD DENSITY ESTIMATION
Wojciech Tarnawski · Kluwer Academic Publishers eBooks · 2006
This paper describes color space clustering based on the multi-field density estimation in color image segmentation process. It is based on the mode seeking approach executed with pdf multi-field estimator calculated from 3Dhistogram or homogram1. It results in clusters found in the color space of the segmented image. The next stage2 merges clusters in the image domain based on the graph's modification. Unlike to many existing clustering algorithms, the proposed approach does not require the knowledge about the number of the color clusters to be generated. Moreover, the clustering result is controlled by the cluster-validity measure, which makes possible correct adjusting of the parameters for the proposed estimation.