基于变分水平集方法的图像分割

发布时间：2018-05-26 09:43

本文选题：变分水平集 + 图像分割　；参考：《中北大学》2017年硕士论文

【摘要】：图像分割是一项应用广泛的图像处理技术,可很大程度的减少后面高级图像处理所需的数据量,且不影响结构特征相关的信息,在图像处理中起关键作用。在图像分割中出现误差将影响后续处理图像的有效性,所以近半个世纪以来,学者们不断提出各种分割方法来提高分割的精度和准确性,在改进分割的方法上做出不少工作。而变分水平集应用在图像分割中也有显著的效果,广泛应用在医学,交通,工业,农业等各种领域。Samson等学者提出将FCM聚类与变分水平集方法相结合进行图像分割,该方法具有很好的图像分割效果。但这种模型需周期性不停地重新初始化,从而影响图像分割时间。本文通过引入内部能量函数的H1正则化,使得水平集函数在演化过程中无需重新初始化从而节约时间。用文中的方法与传统的Samson模型和FCM算法分割图像对比实验,结果表明,本文方法具有更短的运行时间和更好的分割效果。其次,针对传统的CV模型在图像分割中不能很好的分割灰度不均的图像,分析了Lee-Seo和Li-Kim两种改进模型在图像分割方面的性能,提出一种新的能量函数,并给出了一种基于改进的能量函数的图像分割算法。三组实验结果表明,与CV模型、Lee-Seo模型、Li-Kim模型比较起来,改进后的算法具有运行时间短,迭代次数少,分割效果好等优点。最后,对本文做出总结,提出不足与今后可以继续学习研究的方向。
[Abstract]:Image segmentation is a widely used image processing technology, which can greatly reduce the amount of data needed for the subsequent high-level image processing, and does not affect the information related to structural features. It plays a key role in image processing. The error in image segmentation will affect the effectiveness of the subsequent image processing, so in the last half century, scholars have proposed a variety of segmentation methods to improve the accuracy and accuracy of segmentation, and a lot of work has been done to improve the segmentation method. And the application of variational level set in image segmentation also has remarkable effect. It is widely used in medicine, traffic, industry, agriculture and other fields. Samson and other scholars put forward the combination of FCM clustering and variational level set method for image segmentation. This method has good image segmentation effect. However, this model needs to be reinitialized periodically, which affects the time of image segmentation. By introducing the H _ 1 regularization of the internal energy function, the level set function does not need to be reinitialized in the evolution process, thus saving time. Compared with the traditional Samson model and FCM algorithm, the experimental results show that the proposed method has shorter running time and better segmentation effect. Secondly, aiming at the fact that the traditional CV model can not segment the image with uneven grayscale, the performance of the two improved models, Lee-Seo and Li-Kim, in image segmentation is analyzed, and a new energy function is proposed. An image segmentation algorithm based on improved energy function is presented. The results of three groups of experiments show that compared with the CV model Lee-Seo model and Li-Kim model, the improved algorithm has the advantages of shorter running time, fewer iterations and better segmentation effect. Finally, this paper makes a summary, proposes the insufficiency and may continue to study the research direction in the future.
【学位授予单位】：中北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】