| Popis: |
A new method for automated road extraction in remote sensing images is proposed based on Nonsubsampled Contourlet Transform (NSCT). Due to the advantages of multi-scale, multi-direction and translation invariance, NSCT is much efficient for the capture of linear singularity of road target. A pixel-wised feature extraction algorithm is given first based on the analysis of the NSCT coefficient responses for the road target. Final results can be achieved by the simple fuzzy C-means clustering and the postprocessing steps. Experimental results demonstrate that our method outperforms other methods such as Gabor wavelet in accuracy, localization and false detection. |
