针对应用于图像分割的自适应正交投影分解方法的具体算法的实现问题,提出了一种基于两阶段搜索的实现算法。该算法首先通过粗略搜索得到图像直方图中高斯分量的均值和标准差的大致取值范围,再采用0.618法精确搜索来得到准确的均值;然后将直方图投影到若干合适尺度的基函数上来得到多个匹配值,再利用曲线拟合方法估计出分量的标准差和权值,进而计算出最优阈值,以用于图像分割。理论分析和实验结果表明,基于两阶段搜索自适应正交投影分解的图像分割算法不但计算开销少,而且对图像直方图的描述具有良好的性能,并可以获得较好的图像分割效果。

In this paper,a two-stage search fast algorithm for adaptive projective decomposition,which is used in image segmentation,is presented.In the proposed approach,the first stage is to find the approximate values of the mean and the standard deviation of the Gaussian element in image histogram.And the second stage is to search the exact mean value by 0.618 algorithm.Having the exact mean,some projection values are computed by projecting the histogram to some appropriate Gaussian basis functions,and then the standard deviation and the power can be estimated well based on the least-square curve fitting analysis.Finally,the optimal thresholds between different regions can be determined.The theoretical analysis and the comparison experiment show that the proposed approach is more efficient and effective.