摘要
针对低信噪比图象的边缘提取问题,本文提出了一种非线性方法,即利用大窗口平滑去噪性能强与小窗口提取边缘性能好相结合的方法,此时采取大窗口滤波,小窗口中作非线性的微分算子求导。为了避免求导后阈值选取的盲目性,文中提出了一种噪声引导的阈值确定准则,并根据这个阈值分割图象。在大窗口滤波中,采用了“二维卷积等于两个一维卷积级联”的技术压缩滤波器的存储空间。最后对这种方法进行了性能评价,并且给出了实验结果。
A nonlinear method for detecting the edge of low-SNR image is developed. This method adopts the filtering in large window to smooth the noise and the nonlinear diffe-rential operator in small window to detect the edges. A criterion of noise-guided threshold selection is introduced to segment the derivative image so that the threshold can be determined automatically. In large window's filtering, the technique of 2-D convolution implementation by two 1-D convolutions in series is taken to reduce the storage space of the filter. Finally, the performance of this method is evaluated, and experimental results are given.
关键词
图象处理
边缘提取
信噪比
计算机
Image processing
Edge detection
Nonlinear defferential operator
Exponential filter
