The Man-made Areas Segmentation Based on Region Growing Method Using Local Fractal Dimension

摘要

This paper presents a method using the modified region growing method and threshold filter process for the segmentation of man-made areas. The modified region growing method integrates local fractal dimension and region merging process into the traditional region growing algorithm. The proposed 5 × 5 to 3 × 3 method of estimating the local fractal dimension is more stable and accurate besides exact extraction of the local information of images. It is discussed that the region merging process is significant to segment the target areas. Finally, the threshold filter method is used to realize the segmentation of man-made areas and the effectiveness of the proposed method is confirmed through computer simulations.