图像分形维数计算及其边缘提取

Images Fractal Dimension Calculation and Its Edge Extraction

为提高遥感影像中地物边缘信息的提取精度,以离散分形布朗随机场(DFBR:Discrete Fractal BrownRandom field)模型为依据,尝试设计并利用Matlab编程实现一种基于遥感影像单个像元的分形维数计算算法。该算法将影像的灰度空间映射成分形维数空间,然后在该空间进行变换和边缘检测。地物空间分布及其影纹结构边缘特征的差异,使计算分形维数所选窗口大小成为关键。选取研究区局部地段高空间分辨率遥感影像作为实验数据,通过计算不同窗口下像元分形维数,得到最佳边缘信息提取的计算窗口。实验结果表明,该算法在同类计算中更符合遥感数据的特点,提高了遥感影像地物边缘信息提取精度。

Edge information in remote sensing image provides a rich content for image understanding. An algo- rithm to calculate the fractal dimension of a single pixel by using the model of DFBR （ Discrete Fractal Brown Random field） is designed. This algorithm is programmed by Matlab. Map the image gray level space to the fractal dimension of space, and then transform and edge detection in the latter one. Spatial distribution of surface features and the differences of image texture structure and edge features, make the selected window size used to calculate fractal dimension as the key. Select part of the high spatial resolution remote sensing image in the study area as the experimental data. By calculating the fractal dimension of pixels in different windows, the calculation window of the best edge detection results is required. The results show that this algorithm is more suitable for re- mote sensing data. And the accuracy of his remote sensing image extraction is improved.