A Metric Approach to Vector-Valued Image Segmentation

We address the issue of low-level segmentation of vector-valued images, focusing on the case of color natural images. The proposed approach relies on the formulation of the problem in the metric framework, as a Voronoi tessellation of the image domain. In this context, a segmentation is determined by a distance transform and a set of sites. Our method consists in dividing the segmentation task in two successive sub-tasks: pre-segmentation and hierarchical representation. We design specific distances for both sub-problems by considering low-level image attributes and, particularly, color and lightness information. Then, the interpretation of the metric formalism in terms of boundaries allows the definition of a soft contour map that has the property of producing a set of closed curves for any threshold. Finally, we evaluate the quality of our results with respect to ground-truth segmentation data.     

A new active contour remote sensing river image segmentation algorithm inspired from the cross entropy

The CV (Chan–Vese) model is a piecewise constant approximation of the Mumford and Shah model. It assumes that the original image can be segmented into two regions such that each region can be represented as constant grayscale value. In fact, the objective functional of the CV model actually finds a segmentation of the image such that the within-class variance is minimized. This is equivalent to the Otsu image thresholding algorithm which also aims to minimize the within-class variance. Similarly to the Otsu image thresholding algorithm, cross entropy is another widely used image thresholding algorithm and it finds a segmentation such that the cross entropy of the segmented image and the original image is minimized. Inspired from the cross entropy, a new active contour image segmentation algorithm is proposed. The region term in the new objective functional is the integral of the logarithm of the ratio between the grayscale of the original image and the mean value computed from the segmented image weighted by the grayscale of the original image. The new objective functional can be solved by the level set evolution method. A distance regularized term is added to the level set evolution equation so the level set need not be reinitialized periodically. A fast global minimization algorithm of the objective functional is also proposed which incorporates the edge term originated from the geodesic active contour model. Experimental results show that, the algorithm proposed can segment images more accurately than the CV model and the implementation speed of the fast global minimization algorithm is fast.     

A Part-aware Surface Metric for Shape Analysis

The notion of parts in a shape plays an important role in many geometry problems, including segmentation, correspondence, recognition, editing, and animation. As the fundamental geometric representation of 3D objects in computer graphics is surface-based, solutions of many such problems utilize a surface metric, a distance function defined over pairs of points on the surface, to assist shape analysis and understanding. The main contribution of our work is to bring together these two fundamental concepts: shape parts and surface metric. Specifically, we develop a surface metric that is part-aware. To encode part information at a point on a shape, we model its volumetric context – called the volumetric shape image (VSI) – inside the shape's enclosed volume, to capture relevant visibility information. We then define the part-aware metric by combining an appropriate VSI distance with geodesic distance and normal variation. We show how the volumetric view on part separation addresses certain limitations of the surface view, which relies on concavity measures over a surface as implied by the well-known minima rule. We demonstrate how the new metric can be effectively utilized in various applications including mesh segmentation, shape registration, part-aware sampling and shape retrieval.     

A novel active contour model for image segmentation using local and global region-based information

In this paper, we propose a novel level set geodesic model for image segmentation. In our model, we define a hybrid signed pressure force (SPF) function integrating local and global region-based information to segment inhomogeneous images. The local region-based SPF utilizes mean values on local circular regions centered in each pixel. By introducing the local image information, the images with intensity inhomogeneity can be effectively segmented. In order to reduce the dependency on complex initialization, we incorporate a global region-based SPF into this model to develop a hybrid SPF. The global SPF and the local SPF are adaptively balanced by an adaptive weight. In addition, we also extend this model to four-phase level set formulation for brain MR image segmentation. Finally, a truncated Gaussian kernel is used to regularize the level set function, which not only regularizes it but also removes the need for computationally expensive re-initialization. Experimental results indicate that the proposed method achieves superior segmentation performance in terms of accuracy and robustness.     

Finsler active contours

In this paper, we propose an image segmentation technique based on augmenting the conformal (or geodesic) active contour framework with directional information. In the isotropic case, the Euclidean metric is locally multiplied by a scalar conformal factor based on image information such that the weighted length of curves lying on points of interest (typically edges) is small. The conformal factor which is chosen depends only upon position and is in this sense isotropic. While directional information has been studied previously for other segmentation frameworks, here we show that if one desires to add directionality in the conformal active contour framework, then one gets a well-defined minimization problem in the case that the factor defines a Finsler metric. Optimal curves may be obtained using the calculus of variations or dynamic programming based schemes. Finally we demonstrate the technique by extracting roads from aerial imagery, blood vessels from medical angiograms, and neural tracts from diffusion-weighted magnetic resonance imagery.     

Streamline embedding for 3D vector field exploration

We propose a new technique for visual exploration of streamlines in 3D vector fields. We construct a map from the space of all streamlines to points in IR(n) based on the preservation of the Hausdorff metric in streamline space. The image of a vector field under this map is a set of 2-manifolds in IR(n) with characteristic geometry and topology. Then standard clustering methods applied to the point sets in IR(n) yield a segmentation of the original vector field. Our approach provides a global analysis of 3D vector fields which incorporates the topological segmentation but yields additional information. In addition to a pure segmentation, the established map provides a natural "parametrization” visualized by the manifolds. We test our approach on a number of synthetic and real-world data sets.     

基于非均匀热扩散的交互式图像分割算法

交互式图像分割是图像分割中的重要分支,在现实生活和医学领域都有着广泛的应用。该文基于计算测地距离的热方法,引入了热扩散系数,提出了一种基于非均匀热扩散的交互式图像分割算法。该算法利用图像的颜色信息构造三角网格作为热扩散的媒介,首先由热方程找到距离增加的方向,再利用泊松方程还原测地距离。将前景中人工交互区域上的热流扩散速度增加,则前景不同部分之间的测地距离变小,消除了内部边界,通过设置外部边界分割限制条件,即可实现完整的前景分割。算法仅需求解两个稀疏线性方程组,鲁棒性强、精度高且更易于操作。同时,拉普拉斯算子和梯度算子的预计算可以被多次重用,减少了内存占用和时间消耗。大量交互式图像分割实验结果表明:该算法无需过多的用户交互信息,即可将现实图像中的复杂前景快速准确地分割出来。     

融合区域和测地线的活动轮廓模型与图割相结合的自然图像分割

针对活动轮廓模型利用水平集函数演化来分割图像时,只能分割灰度均匀的图像 问题以及容易陷入能量泛函局部极小值的缺点,提出一种新的图像分割模型。模型将区域中的 局部和全局信息融合的活动轮廓模型与边界模型相结合,然后利用图切割进行优化。实验表明, 该方法对初始曲线不敏感,能分割灰度不均的自然图像,避免陷入局部极小,并能有效提高图 像分割的速度和精度。     

无需重新初始化的自适应快速水平集演化模型

水平集方法已被广泛地应用在图像分割中,传统的水平集方法需要通过周期性的初始化水平集函数使得它一直保持在符号距离函数附近,然而初始化与水平集理论和实现相违背。最近,Li C等人提出一种完全不需要初始化的变分模型,该模型的主要不足就是单方向演化,即演化曲线或收缩或扩张到目标边界。针对二值图像提出一种新的基于距离保持水平集方法的活动轮廓模型,它不依赖于初始位置,演化曲线准确地收敛在目标边界,更重要的是曲线演化只需一次迭代。     

结合分布度量统计建模的主动轮廓图像分割

图像分割是数字图像处理中不可或缺的关键步骤。为了解决传统主动轮廓模型针对非匀质图像分割结果不准确且分割效率低的问题,提出一种结合分布度量统计建模的主动轮廓图像分割算法。所提算法的能量驱动力兼顾了图像的全局统计建模信息和其他混合灰度分布信息,使得分割曲线能够更加精确地演化至目标边缘。分布度量能量驱动力定义为轮廓内外概率密度函数定义的比率距离的方差,该能量驱动力基于图像全局信息统计建模,能够更加精确地描述轮廓曲线内外的能量变化;混合灰度分布能量驱动力由图像灰度值与融合均值与中值的区域拟合中心的L2范数表示。将分布度量能量驱动力与混合灰度分布能量驱动力组合形成新的能量泛函,利用水平集方法和梯度下降法迭代求得该能量泛函的最小值,以获得最终的图像分割结果。与传统CV（Chan Vese）模型、LBF（Local Binary Fitting）模型等四种算法的图像分割结果相比,所提模型在主观视觉效果、对初始轮廓的敏感性、运行时间和迭次次数方面均具有较大优势。     

Learning a color distance metric for region-based image segmentation

In this paper we describe an experiment where we studied empirically the application of a learned distance metric to be used as discrimination function for an established color image segmentation algorithm. For this purpose we chose the Mumford–Shah energy functional and the Mahalanobis distance metric. The objective was to test our approach in an objective and quantifiable way on this specific algorithm employing this particular distance model, without making generalization claims. The empirical validation of the results was performed in two experiments: one applying the resulting segmentation method on a subset of the Berkeley Image Database, an exemplar image set possessing ground-truths and validating the results against the ground-truths using two well-known inter-cluster validation methods, namely, the Rand and BGM indexes, and another experiment using images of the same context divided into training and testing set, where the distance metric is learned from the training set and then applied to segment all the images. The obtained results suggest that the use of the specified learned distance metric provides better and more robust segmentations, even if no other modification of the segmentation algorithm is performed.     

一种改进的变分水平集的图像分割算法

为了解决灰度不均匀现象对医学图像的干扰问题,提出了基于局部极性信息的活 动轮廓模型。通过引入局部图像信息,该模型能有效地分割灰度不均匀图像。在规则化项中增 加的能量惩罚项,使得水平集函数在演化过程中保持为近似的符号距离函数。该算法将图像分 割问题归结为曲线能量泛函的最小化,首先建立包含局部灰度信息（极性信息）和改进的符号 距离函数的曲线演化能量泛函;然后采用变分水平集方法求解能量函数的最小值,得到最终的 分割结果。真实医学图像和人工合成图像的实验结果表明,此方法对灰度不均匀的医学图像有 较高的分割精确度,在图像分割速度上有较大提高。由于利用了局部灰度信息,可以有效地分 割灰度不均匀的医学图像,而改进后的变分水平集可以完全避免重新初始化,使得图像分割效 率大大提高了。     

On the Geometry of Multivariate Generalized Gaussian Models

This paper concerns the geometry of the zero-mean multivariate generalized Gaussian distribution (MGGD) and the calculation of geodesic distances on the MGGD manifold. The MGGD is a suitable distribution for the modeling of multivariate (color, multispectral, vector and tensor images, etc.) image wavelet statistics. Expressions are derived for the Fisher-Rao metric for the zero-mean MGGD model. A closed-form expression is obtained for the geodesic distance on the submanifolds characterized by a fixed MGGD shape parameter. Suitable approximate solutions to the geodesic equations are presented in the case of MGGDs with varying shape parameters. An application to image texture similarity measurement in the wavelet domain is briefly discussed, comparing the performance of the geodesic distance and the Kullback-Leibler divergence.     

融合语义与几何特征的人体模型结构分割*

针对现有三维人体模型形状分析方法存在人工干预及对姿势依赖的问题,提出一种融合语义与几何特征的三维人体形状分析方法。首先,基于模型表面测地线距离以及内部空间体积特征的度量,提出了基于骨架树的结构检测方法;其次,基于人体测量学先验语义知识,进一步提炼模型的层次结构。该方法能有效的提取不同姿态人体模型的结构特征,并实现基于语义的模型分割,一系列实验结果验证了该方法的高效性与鲁棒性。     

基于核特征距离的局部活动轮廓模型

提出了一种基于核特征距离局部活动轮廓分割模型。在模型中使用核特征距离来构造局部拟合能量,从而可以获取精确的局部图像特征,可以分割存在灰度不均匀的图像。并通过引入水平集规范项以避免水平集演化的重新初始化,提高了分割的效率。实验结果表明,本模型可以很好地克服灰度不均匀性,同时在分割精度上有了较大的提升,特别是分割速度比LBF模型快1.3～1.5倍。     

基于改进符号压力函数的变分水平集图像分割算法

为了更好地解决含有弱边界、灰度不均匀的图像在分割时出现的轮廓线错误移动而导致分割结果错误的问题,结合图像的统计信息,构造出一种新的符号压力(SPF)函数,提出了一种基于改进的压力符号函数的变分水平集图像分割算法。首先,利用新的压力符号函数代替边缘函数,构造了新的活动轮廓模型;其次,该算法保持了测地线活动轮廓(GAC)模型和chan-vese(C-V) 模型的优点,使水平集函数演化到目标的边界上;最后,对一些弱边界、灰度不均匀的图像进行仿真实验,结果表明提出的算法能够精准地分割目标,并且具有一定的抗噪性。     

基于区域中心的交互式图像前景提取方法

图像分割是从图像中提取有意义的区域,是图像处理和计算机视觉中的关键技术。而自动分割方法不能很好地处理前景复杂的图像,对此提出一种基于区域中心的交互式图像前景提取算法。针对图像前景的复杂度,很难用单一的相似区域描述前景,文中采用多个区域中心来刻画目标区域。为提升图像分割的稳定性,给出基于超像素颜色、空间位置和纹理信息的相似性度量方法;为确保图像分割区域的连通性和准确性,定义了基于超像素的测地距离计算方法。使用基于测地距离的超像素局部密度,来分析图像的若干区域中心;基于用户交互的方式来分析前景的区域中心,得到图像前景。经过大量彩色图像的仿真表明,在分割过程中利用少量的用户交互信息,可有效提升图像分割的稳定性和准确性。     

基于变分水平集的图像模糊聚类分割

结合变分水平集方法和模糊聚类,提出了一个基于变分水平集的图像聚类分割模型.该模型引入了一个基于图像局部信息的外部模糊聚类能量和一个新的关于零水平集的正则化能量,使得该模型对噪声图像的聚类分割更具鲁棒性.通过在能量泛函中加入一个内部约束能量约束水平集函数为符号距离函数,可以使水平集演化过程无需重新初始化.进一步提出了一种变分形式的聚类中心更新方法,实现了半监督的图像聚类分割.实验中采用不同类型的图像与FCM聚类模型、CV模型、Samson模型进行了对比实验,实验结果显示,该模型能够克服图像中噪声的影响,取得较满意的聚类分割效果.     

结合显著性和边缘信息的水平集图像分割方法

针对 LBF 模型对初始轮廓的依赖性和对边缘的弱控制能力,研究了一种结合显著性和边缘信息 的水平集图像分割方法。首先,结合小波分析理论,基于视觉注意机制构造图像显著图;然后,利用小波分解 所描述的图像边缘信息,构造边缘检测函数,同自适应初始轮廓一起引入到 LBF 水平集模型中,并用有限差 分法进行数值求解。实验结果表明,提出的图像分割方法能有效降低初始轮廓位置对活动轮廓模型的影响,对 合成图像、自然图像均有较好的分割结果,相较于其他传统方法具有更高的演化效率和分割质量。     

Fast level set image and video segmentation using new evolution indicator operators

We propose an effective level set evolution method for robust object segmentation in real images. We construct an effective region indicator and an multiscale edge indicator, and use these two indicators to adaptively guide the evolution of the level set function. The multiscale edge indicator is defined in the gradient domain of the multiscale feature-preserving filtered image. The region indicator is built on the similarity map between image pixels and user specified interest regions, where the similarity map is computed using Gaussian Mixture Models (GMM). Then we combine these two methods to develop a new mixing edge stop function, which makes the level set method more robust to initial active contour setting, and forces the level set to evolve adaptively based on the image content. Furthermore, we apply an acceleration approach to speed up our evolution process, which yields real time segmentation performance. Finally, we extend the proposed approach to video segmentation for achieving effective target tracking results. As the results show, our approach is effective for image and video segmentation and works well to accurately detect the complex object boundaries in real-time.