摘要
提出了一种基于多尺度形态学融合的分水岭图像分割方法,采用多尺度结构元素对输入图像并行滤波,并对结果图像进行基于小波变换的图像信息融合,针对小波分解的不同频率域,选择不同的融合规则,既抑制噪声又保持目标轮廓信息。最后,采用最大熵法自适应确定算法的初始阈值,并给出一种有效的区域合并方法来优化分割结果。实验证明,采用此分割方法可以获得较好的分割结果。
The image segmentation based on the watershed method always results in over-segmentation. The novel watershed segmentation algorithm is proposed to solve this problem. The target image is filtered using multiscale structure element with the parallel processing. This paper presents the image fusion using the wavelet transform and chooses fusion principles, which can eliminate the noise while preserving the main contour of target. Finally, the image maximum entropy is adopted to determine the initial threshold of the watershed and segmentation results are improved by the effective region merging method. Experiments show that the proposed algorithm is efficient.
出处
《数据采集与处理》
CSCD
北大核心
2006年第4期398-402,共5页
Journal of Data Acquisition and Processing
基金
国家高技术研究发展计划("863"计划)(2004A783052)资助项目
关键词
分水岭
多尺度
小波变换
图像融合
图像分割
watershed
multiscale
wavelet transform
image fusion
image segmentation
