摘要
在对二值化方法 Otsu算法分析的基础上,提出一种差分算子与改进Otsu算法相结合的新算法。该算法通过差分算子保留原图的边界特征,然后再搜寻出与两类类内均值的平均值,并找出该平均值整数部分相等的阈值,确定一个符合Otsu准则的阈值,,然后将一个大的图像分割成若干小的块进行二值化。实验结果表明,该算法能够较好地保留原图的边界信息,有效地提高了低质量图像识别准确率。
Based on binarization method of Otsu algorithm,a new methodology of combination of difference operator and improved Ostu was proposed. The boundary characteristics of the original image were retained by difference operator,and the mean of two classes partitioned was searched. The threshold equal to the integer part of the mean value can be found,thereby the threshold according to the Otsu rules was got,eventually a big image can be divided into a large amount of small blocks to binarize. The experimental results implicate that the approach can keep the boundary information from the original image well,which effectively improves the correct rate ofidentification of low quality images.
引文
[1]OTSU N.A threshold selection method from gray-level his to grams.IEEE Transactions on System Man and Cybernetic,1979,9(1):62-66.
[2]邵立康,邹飞平.一种基于直方图的阈值分割算法.CT理论与应用,2009,18(2):66-77.
[3]汪荣贵.一种新的自适应二维Otsu图像分割算法研究.中国科学技术大学学报,2010,40(8):842-846.
[4]张新明.二维直方图准分的Otsu图像分割及其快速实现.电子学报,2011,39(8):1778-1784.
[5]许向阳,宋恩民.Otsu准则的阈值性质分析.电子学报,2009,37(12):2716-2719.
[6]王强,马利庄.图像二值化时图像特征的保留.计算机辅助设计与图形学学报,2000,12(10):746-749.
[7]陈强,朱立新.结合Canny算子的图像二值化.计算机辅助设计与图形学学报,2005,17(6):1302-1305
[8]何志勇,孙立宁.Otsu准则下分割阈值的快速计算.电子学报,2013,41(2):267-272.
[9]FAN J L,ZHAO F.Two dimensional Otsu.s curve thresh-olding segmentation method for gray levelimages.Acta Electronica Sincia,2007,35(4):751-755.
[10]汪海洋,潘德炉,夏德深.二维Otsu自适应阈值选取算法的快速实现.自动化学报,2007,33(9):968-971.