快速二维直方图斜分最小误差的图像阈值分割

Fast Image Thresholding with Minimum Error Based on 2-D Oblique Segmentation

鉴于二维斜分法的优势,提出了一种快速二维直方图斜分最小误差的阈值分割方法。首先将二维直方图斜分原理运用到最小误差阈值法中使得分割更准确,然后对其阈值选取公式进行简化得到最简公式,并利用此公式导出其一般递推算法,最后将二维直方图概率分布特性与这种算法有机结合得到新型的递推算法来提高运行速度。实验结果表明,与二维直线型最小误差阈值分割法相比,算法效率更高,与其递推算法相比,所提出的新型递推算法的运行速度更快,约快4倍。

In view of the advantages of the thresholding minimum error thresholding method based on 2 - D method based on 2 - D oblique segmentation, a fast histogram oblique segmentation is proposed in this paper. First the thresholding method based on minimum error was formulated using the approach based on two-dimensional histogram oblique segmentation to segment images more accurately. Then, the 2 - D minimum error thresholding formula was simplified, and its recursive algorithm of 2 - D histogram oblique segmentation was deduced on the simplified formula. Finally, the probability distribution feature of 2 - D histogram and the algorithm were combined to get a novel recursive algorithm in order to improve the running speed. Experimental results show that compared with the current two-dimensional linear-type minimum error threshold segmentation method and its recursive algorithm, the proposed algorithm is more efficient and the novel recursive algorithm＇s running speed is about four times faster.