Segmentation Approach for Remote Sensing Images Based on Local Homogeneity Gradient and Its Evaluation
基于局域同质性梯度的遥感图像分割方法及其评价

A fast segmentation approach for multi-band remote-sensing images was proposed. First, remote-sensing images were quantized. Second, local homogeneity gradient images were produced based on quantization results. Third, a fast watershed transform was applied to gradient images to perform initial segmentation. Finally, an efficient region merging approach was implemented to obtain final segmentation results. To validate the proposed approach, experiments on Quickbird images were performed, and segmentation evaluations were conducted. Evaluation results indicated that the proposed approach was a suitable one for segmenting remote-sensing images, which is better than the Muhiresolution segmentation approach with respect to segmentation speed and accuracy.