基于全局和局部相关熵活动轮廓模型的超声图像分割算法

Ultrasound image segmentation algorithm based on global and local correntropy K-mean active contour model

局部相关熵K均值(LCK)模型在非高斯噪声和图像灰度不均匀时具有较好的图像分割效果,但是其计算复杂度较高,收敛较慢。针对该问题,本研究将局部相关熵能量项和全局相关熵能量项结合,提出了全局与局部相关熵K均值(GLCK)图像分割算法。其中局部相关熵力在目标边界附近起主导作用,用来吸引水平集函数曲线到达目标边界,而全局相关熵力在远离目标边界处起主导作用。对超声医学图像和人工合成图像进行了图像分割实验,并同LCK等模型进行了对比,结果表明提出的GLCK模型具有更好的鲁棒性和图像分割精度,并且计算时间也显著减少。另一方面GLCK模型对于噪声和模糊边界影响严重的超声医学图像,具有好的分割效果。

The local correntropy K-mean（ LCK） model has a better image segmentation effect when the non-Gaussian noise and the image grayscale inhomogeneous,but the computational complexity is rather high and the convergence is rather slow. In order to solve this problem,a global and local correntropy K-mean（ GLCK） image segmentation algorithm was proposed by combining the local correntropy and the global correntropy energy. The local correntropy force played a leading role near the target boundary,which was used to attract the level set function curve to reach the target boundary,and the global correntropy force played the leading role at the distance from the target boundary. The image segmentation experiments of ultrasonic medical images and synthetic images were carried out and compared with the LCK and other models. The results show that the proposed GLCK model has better robustness and image segmentation precision,and the calculation time is also significantly reduced. On the other hand,the GLCK model has a good segmentation effect for ultrasound medical images with severe noise and blurred boundaries.