融合RSF模型及边缘检测LOG算子的图像分割方法的研究

Image segmentation method based on RSF model and edge detection LOG operator

针对强度不均的图像,采用可变区域拟合能量模型(RSF)进行分割效果较好,但这取决于初始轮廓的位置,初始轮廓选择不好,直接导致分割结果出错。文中提出了一种将自适应可变区域拟合(Adaptive RSF)能量和优化的二阶微分边缘检测算子(LOG)结合的活动轮廓模型,进行图像分割。首先,计算闭合曲线内外信息熵,使其RSF模型自适应调整权值;其次,提出一个能量泛函优化LOG项,它可以平滑同质区域,同时增强边缘信息;然后,将优化后的LOG能量项与ARSF能量项结合起来,利用局部区域信息将曲线驱动到边界,在LOG项的加入下,实现了初始轮廓的自由设定,实现轮廓精确提取;最后,通过单个细胞图像进行实验,证明该模型不仅具有良好的鲁棒性,而且具有更高的分割精度和效率。

For the image with uneven intensity, the segmentation is effective by using the variable region scalable fitting （RSF） energy model, but it depends on the position of the initial contour. The initial con- tour selection is not good and leads directly to the segmentation result. This paper proposes an adaptive contour model by combining adaptive RSF （ARSF） energy with the optimized LOG energy to conduct im- age segmentation. Firstly,an automatic adjustment of the weights of the RSF model is realized based on the entropy of the closed curve. Secondly, an energy functional optimization LOG item is used to smooth the homogeneous region and enhance the edge information. Then, combined the optimized LOG energy term with the ARSF energy item, the local area information is used to drive the curve to the boundary. With the addition of the LOG item,the initial contour can be freely set and the outline can be accurately extracted. Finally,the experiments with a single cell image demonstrate that the model has good robust- ness, and the higher segmentation accuracy and the efficiency.