摘要
为提高被噪声、纹理等干扰的图像分割精度,提出了一种基于Bayesian最小误判的活动轮廓模型.首先将目标和背景这两类中心的马氏距离作为误判概率的判决规则,利用方向扩散PDE方程对观测图像进行演化,使得扩散只沿切线方向发生,得到原图像的一个平滑近似图像,从而使目标和背景的马氏距离增大,误判概率减小.然后结合局部二值拟合活动轮廓模型,对特征函数进行迭代演化,直至收敛到目标边缘.最后采用结合梯度下降法和迭代卷积阈值化方法对模型进行数值求解.实验结果表明,提出的模型可以很好地排除干扰信息的影响,准确地提取目标轮廓,误判概率更小.和几个经典变分图像分割模型相比,DSC、IOU等指标值都表明提出模型具有较好的实验效果.
In order to improve the segmentation accuracy of images disturbed by noise and texture,an active contour model based on Bayesian minimum misjudgment is proposed.Firstly,the Mahalanobis distance between the target and the background is used as the decision rule of the misjudgment probability,and the PDE equation of directional diffusion is used to evolve the observed image,so that the diffusion occurs only along the tangent direction,and a smooth approximate image of the original image is obtained,thus the Mahalanobis distance between the target and the background increases and the misjudgment probability decreases.Then,the active contour model fitting is conducted in combination with local binaries,and the feature function is iteratively evolved until it converges to the target edge.Finally,the model is solved numerically by the method of gradient descent and iterative convolution thresholding.The experimental results show that the proposed model can well eliminate the influence of interference information,accurately extract the target contour,and reduc the probability of misjudgment.Compared with several classical variational image segmentation models,the values of DSC and IOU indicate that the proposed model has better experimental results.
作者
党红宇
唐利明
徐雅雅
DANG Hongyu;TANG Liming;XU Yaya(School of Mathematics and Statistics,Hubei Minzu University,Enshi,Hubei 445000,China)
出处
《内江师范学院学报》
CAS
2024年第4期53-64,共12页
Journal of Neijiang Normal University
基金
国家自然科学基金资助项目(62061016,61561019)。