变分法图像分割算法研究

发布时间：2019-01-28 17:59

【摘要】：变分法之所以成为图像分割理论研究的热点问题是因其建模过程比较容易、方法的扩展性较好、实现过程简单,已成为图像分割处理范围中较为有效且广泛使用的方法之一。针对图像分割算法目前所存在的问题,本文意在提高分割结果的准确度和鲁棒性,降低图像分割的时间复杂度,对使用变分水平集法进行图像的分割处理做了较深入的探讨,研究两种变分方法实现图像分割。由于传统的活动轮廓存在对图像分割鲁棒性欠好的问题,本文将区域相关的轮廓模型(Active Contours without Edge,CV模型)和梯度相关的轮廓模型(Geodesic Active Contour,GAC模型)结合使用,通过图像熵与图像的梯度和将两部分进行结合。首先,将图像熵与图像梯度和拟合为权重因子,构建基于图像梯度的轮廓模型和基于图像区域的轮廓模型相结合的分割模型。其次,将水平集函数方法应用到模型中,这样模型在分割过程能够连续,也可以适应拓扑变化。最后,为了提升曲线演化的速度,论文采用窄带方法使活动曲线能够快速演化。通过实验证明,使用该模型方法分割图像,其鲁棒性较好,同时分割速度也较快。传统活动轮廓分割图像时间复杂度有时不甚理想,本文设计了基于轮廓邻域梯度进行拟合的活动轮廓模型。该模型通过计算轮廓线邻域梯度值,并用计算出的梯度值作为阈值来拟合迭代函数,从而减小分割图像的时间复杂度。通过实验能够证明,本文算法能够有效快速地分割图像,并有较小的时间复杂度。两种改进的变分技术能够较好地提高图像分割质量,为图像分割技术在生物工程、医疗领域和工业设计等诸多方面的应用提供技术支持。
[Abstract]:Variational method has become one of the most effective and widely used methods in image segmentation because its modeling process is easy, its expansibility is better, and its realization process is simple. Aiming at the existing problems of image segmentation algorithm, this paper aims to improve the accuracy and robustness of segmentation results and reduce the time complexity of image segmentation. Two variational methods for image segmentation are studied. Because the traditional active contour is not robust to image segmentation, the region-dependent contour model (Active Contours without Edge,CV model) and gradient dependent contour model (Geodesic Active Contour,GAC model) are used in this paper. The image entropy is combined with the gradient of the image and the two parts are combined. Firstly, image entropy and image gradient fitting are used as weight factors, and a segmentation model based on image gradient and image region based contour model is constructed. Secondly, the level set function method is applied to the model, so that the model can be continuous in the segmentation process and can also adapt to the topological changes. Finally, in order to improve the speed of curve evolution, the narrow band method is used to make the activity curve evolve rapidly. Experimental results show that the proposed method is robust and fast. The time complexity of traditional active contour segmentation is not ideal. In this paper, an active contour model based on contour neighborhood gradient fitting is designed. By calculating the contour neighborhood gradient value and using the calculated gradient value as the threshold to fit the iterative function, the model reduces the time complexity of image segmentation. Experimental results show that the proposed algorithm can segment images efficiently and quickly, and has less time complexity. The two improved variational techniques can improve the image segmentation quality and provide technical support for the application of image segmentation technology in bioengineering, medical field and industrial design.
【学位授予单位】：长春工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】