摘要
分水岭分割方法虽然是一种有效且常用的图像分割方法,但它容易产生过分割现象且对噪声非常敏感。为克服这一问题,文中结合小波变换提出了一种改进的分水岭算法。首先利用小波分解变换对形态学梯度图像进行二层分解去噪,通过设置阈值向量对高频小波系数进行阈值处理,重构二维小波;利用形态学标记前景和背景的技术,结合重构的二维小波,得到新的仅在前景和后景标记位置有极小值的分割函数。最后,在修改后的梯度图像上进行分水岭变换,从而取得良好的图像分割效果。
Watershed segmentation method is an effective and commonly used segmentation method of image, but it is easy to cause the over segmentation and very sensitive to noise. In order to overcome this problem,a watershed algorithm based on wavelet transform is proposed in this paper. First,a method of two-layer denoising based on wavelet decomposition transform is used to the morphological gradient image, in which the high-frequency wavelet coefficients is dealt with threshold by setting the threshold vector and the 2-D wavelet is reconstructed. Then, using the technology of marking foreground and background in morphology and composing the 2-D reconstructed wavelet,a modified new segmentation function is obtained and it has only minimal value at the marked position in the foreground and background. At last, watershed transform is used to the modified gradient image and a better effect of segmentation is achieved.
出处
《计算机技术与发展》
2016年第3期108-112,共5页
Computer Technology and Development
基金
国家自然科学基金资助项目(11475135)
关键词
小波变换
分水岭图像分割
图像梯度
小波二层分解去噪
wavelet transform
watershed image segmentation
image gradient
two-layer wavelet decomposition to denoise
