二维直方图θ-划分最小误差图像阈值分割

Image Thresholding Based on 2-D Histogram θ-Division and Minimum Error

针对常用二维直方图区域直分法存在错分的问题,并为适应实际中不同图像及分割目的的需要,提出了更具普适性的二维直方图θ-划分最小误差阈值分割方法(θ为分割直线的法线与灰度级轴的夹角).导出了相应的阈值选取公式及其快速递推算法,根据实验结果分析了θ取值对分割结果和算法运行时间的影响.与二维直方图直分最小误差法相比,所提方法的分割结果更为准确,抵抗噪声更为稳健,且所需运行时间也大为减少;而直线形最小误差法只是文中方法中θ=45°的特例.

Aiming at the problem of wrong segmentation in common 2-D histogram region division, in order to meet the requirement of different images and segmentation objectives, the 2-D linear-type minimum er- ror threshold segmentation method was generalized, and a much more widely suitable thresholding method was proposed based on 2-D histogram θ-division and minimum error. The threshold selection formulae and its fast recursive algorithm were deduced. The influence of different θ values on segmented results and run- ning time was analyzed according to the experimental results. Compared with the conventional 2-D mini- mum error method, the proposed method not only achieves more accurate segmented result and more ro- bust anti-noise, but also significantly reduces the running time. The linear-type minimum error threshold segmentation method is only a special case with θ=45° of the proposed method.