摘要
针对分水岭分割算法的2个缺陷:耗时较长和过分割问题,该算法在低分辨率图像上进行分水岭分割,提高了分割的速度;由低分辨率图像返回到高分辨率图像时,采用了一种基于边缘信息的合并函数,避免了边缘信息的丢失,保证了分割的准确性。该文设计了一种基于梯度图像的噪声抑制方法,可抑制高斯噪声对梯度图像的影响,有效避免了过分割问题。实验结果证明,该算法兼顾了效率和分割的准确性。
This paper proposes a novel watershed segmentation based on multi-resolution image, in order to overcome the drawbacks of traditional watershed segmentation: low calculated efficiency and over-segmentation. Watershed segmentation is completed in low resolution to reduce the burden of computer. A new function based on the edges is proposed to merge regions, which can detect the high frequency information lost in low resolution image. An adaptive threshold is proposed for gradient image to denoise. The arithmetic reduces the effect of Gaussian noise to avoid over-segmentation and reduces the burden of merging. Experiments show that the method balances calculated efficiency and segmentation accuracy.
出处
《计算机工程》
EI
CAS
CSCD
北大核心
2006年第19期202-204,207,共4页
Computer Engineering
关键词
多分辨率
分水岭
小波变换
区域合并
Multi-resolution
Watershed
Wavelet transform
Region merger