摘要
本文提出一种新的二维直方图区域斜分方法,导出了基于二维直方图区域斜分的最小类内方差阈值分割快速迭代算法,在实验结果和分析中给出了分割结果和运行时间,并与基于二维直方图直分的O tsu原始算法及其他三种改进算法进行了比较。结果表明本文提出的二维直方图区域斜分方法可以运用于几乎所有的基于二维直方图的阈值分割,使分割后的图像内部区域均匀,边界形状准确,更有稳健的抗噪性。基于区域斜分的最小类内方差阈值分割快速迭代算法的运行时间与二维O tsu原始算法和文献[12]中的改进算法相比减少了4个数量级,约为区域直分O tsu快速递推算法的1/4,不到文献[11]中快速算法的4%。
A new two-dimensional histogram oblique segmentation method is proposed in this paper. The fast iteratire algorithm formulae of the minimum within-class variance thresholding based on two-dimensional histogram oblique segmentation method are deduced. The thresholding results and processing time are given in the experimental result and analysis, which are compared with those of the original Otsu algorithm based on two-dimensional histogram vertical segmentation and its three improved algorithms. The results show that two-dimensional histogram oblique segmentation could be used in nearly all the two-dimensional histogram thresholding. It makes the inner region of the thresholding image uniform, the edge accurate, and has better tolerance capability to noise. Compared with original two-dimensional Otsu algorithm and the improved algorithm in reference [ 12 ] , the processing time of the proposed fast iterative algorithm based on two-dimensional histogram oblique segmentation is reduced 4 orders of magnitude and is about one quarter of that of the recurring algorithm based on vertical segmentation and less than 4% of that of the fast algorithm in reference [ 11 ].
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2008年第12期2651-2657,共7页
Chinese Journal of Scientific Instrument
关键词
图像处理
阈值分割
二维直方图区域斜分
快速迭代算法
image processing
thresholding
two-dimensional histogram oblique segmentation
fast iterative algorithm
