基于局部分形维数的遥感图像分割

本文针对遥感图像的分形特点,采用局部分形维数进行图像分割.利用金字塔模型替代了局部分形计算中的窗口模板,从而对局部三角棱柱面法进行了改进,减少了图像局部分形维数的运算量.并设计了一种新的递归计算流程,有效地降低了内存存储量.利用此法对遥感图像进行分割,结果表明:改进后的三角棱柱面法提高了基于遥感图像纹理特征的图像分割方法的计算速度,同时保留了三角棱柱面法对于分形维数计算的准确性.

Remote Sensing Image Segmentation Based on Local Fractal Dimension

According to the fractal characters of remote sensing image,the image was segmented with the Local Fractal Dimension (LFD).The pyramid model instead of local window model was used to decrease the computation time in computing local triangular prism.Compared with traditional method,a new recursive procedure was designed to decrease the needed memory capacity.The experiment results on remote sensing image segmentation show that the improved method effectively increases the speed of image segmentation based on the texture,and keeps the accurateness of the fractal dimension with the triangular prism method.