Abstract: This paper proposes a new method to solve the problems of region-based segmentation in high-resolution remote sensing sea images in order to partition ships region from complicated background. It gets some information from wavelet transform in an appropriate proportion, does some segmentation by an improved DBC algorithm, and maps to the original image to segment in a smaller scale. Experimental results prove that the algorithms is effective.

Key words: remote sensing image, wavelet transform, fractal dimension, differential box counting, region-based segmentation

摘要： 针对高分辨率遥感海面图像中的船只检测问题，提出一种新的区域分割方法。该方法利用小波变换得到适合分形分析的一定尺度的频谱信息，针对图像特点采用改进的差分盒维计算方法进行区域分割，根据小波变换的尺度映射关系将得到的目标区域映射到原图的相应区域，同时利用分维数的特点进行更细致的分割。实验结果证明了该方法的有效性。

关键词: 遥感图像, 小波变换, 分维, 差分盒维, 区域分割

CLC Number: 

• TP751