【摘要】 阈值分割是一种重要的图像分割方法,是图像检测与识别的重要预处理步骤之一,在计算机视觉等研究领域中有着非常广泛的应用。对传统二维最大熵阈值分割算法的性能进行了分析,针对算法计算复杂度大和阈值判别存在误差这两个不足,提出了一种改进的最大熵阈值分割方法。改进方法通过引入新的阈值门限选取方案来消除误差,并将原本复杂的二维解空间的求解过程简化到了一维,大大减少了计算量。由于等效引入了一个平滑噪声的预处理过程,新方法还克服了噪声对阈值选取的影响。仿真结果表明:改进新方法不仅简单易行,对低对比度图像的分割效果也较好。更多还原

【Abstract】 Thresholding is an important method of image segmentation,and it is one of the essential pretreatment steps of image sensing and recognition.It has been widely used in the field of computer vision.After analyzing the performance of 2D entropic thresholding,a modified approach of max entropic thresholding is presented for reducing the computational complexity and the errors of choosing the threshold.The proposed method eliminated the errors by introducing a new law of choosing the threshold,and this technique reduced the computational complexity by predigesting the solution space from 2D into 1D at the same time.The new method overcame the affect of noises by introducing a pretreatment process of smoothing the image.Experimental results made it clear that the new method is easy to realized,and the segmentation performance in those low contrast images is good as well.更多还原