基于序贯滤波的水平集图像分割

传统基于边缘的水平集分割模型对非均匀区域分割效果不理想,为了解决这一问题,本文分析了序贯滤波的平滑能力与滤波次数之间的关系,将序贯滤波与水平集分割相结合提出了基于序贯滤波的图像分割模型.为了实现自适应于图像内容的平滑,根据平滑分量分割区域的置信度,设计了图像分割的平滑指标,控制序贯滤波次数,使得不同平滑分量的轮廓曲线收敛于目标边界,改善了传统基于水平集方法对非均匀区域的分割效果.本文分割算法的F测度,精确率和召回率均高于传统模型,在一定程度上提高了非均匀区域的分割效果.

The Level Set Image Segmentation on Sequential Filtering

Abstract: The classical edge-based level set segmentation models are not suitable for inhomogeneous region. To improve this, this paper analyzes the relationship between the sequential filtering and the number of the filter, combines the sequential filtering and the level set, finally, presents a new image segmentation model on the sequential filtering. To make the number of the smoothing depends on image data self-adaptively, this model segments the confidence of the region with smoothing components, designs the smoothing indicator, and controls the number of the sequential filtering. Let the contour curve with different smoothing components converge at the object boundary, effectively improve the results of the classical edge-based level set segmentation on inhomogeneous region. The F-measure, precision and recall of image segmentation in this model are all higher than the classical model, and the segmentation effect of inhomogeneous region is improved to some extent.