Research On Unsupervised Segmentation Method Of Color-Texture Image Based On Multi-Scale Structure Tensor

Image segmentation is a fundamental task in field of digital image processing and vision computing, and it can divide the original image into several non-overlap sub-regions and enclosed smoothly curves so that these segment results contain some special meanings. Consequently, image segmentation is becoming a basis topic for researching. As we known, image segmentation is widely applied in object recognition, scene analysis, special effect in movie, object detection, medical image processing, industrial detection, and content based image retrieval and so on, the reason is that it can provide some vital important cues for high level applications such as tracking, detecting and recognition et.al. However, as we lack the deeply understand about the vision mechanism of human, and the content contained in color-texture is diversity, complexity, and randomly. Thus, there is still not appeared a mature approach to meet with any kinds of need in application. Base on the actual circumstance, image segmentation is becoming a difficult spot and hot topic, and how to improve the segmentation result to achieve the generalized and unified task, is still a challenge and arduous mission, and not tackled idealy until now.Aim at the imitated function of computer, such as micro structure, macro structure, and abstract information captured adaptively, and meanwhile it can further acquires the important semantic information. As a result, these functions lead to the unsupervised segmentation approach is focusing broadly, and it should achieve the segmentation tasks, such as distinguish and discriminate the different object accurately, recently. However, due to the abundant color information such as chrome, saturation, illumination and brightness are contained in natural image. Along with the diversity attributes, such as linearity, homogeneity, regularity, randomly, smoothly, roughness, density, granularity, frequency, phase, and directivity of texture, which are related to circumstance relationship closely. Thus, the integrated method of color information and texture information is focused significantly in our article. Due to the requirement about robustness, vigorously, and instantaneity in application, the research about Graph Cuts optimization methods has become one of the most popular methods in recent years. It has many excellent features, such as with the ability to fuse a wide range of visual cues and constraints, global optima, numerical robustness, practically efficiency, and unrestricted topological properties of weighted graph for segments etc. Nevertheless, the traditional Graph Cut methods are not skilled in applicability, accuracy, robustness, and the real-time, therefore, these factors are severely limited the usability of the Graph Cut methods when applying in color texture image segmentation, such as complex, various, random, and class number uncertain situation. Therefore, to alleviate the key problems mentioned above, we carry out a qualitative analysis and quantitative research, theoretical and practical research accordingly, then, several new unsupervised color-texture segmentation approaches are proposed. Concretely, the main innovation and research achievements of this dissertation can be described as follows:Firstly, a new color-texture descriptor was constructed by integrating the compact multi-scale structure tensor (MSST) texture, RGB color information, and the total variation flow (TV). As the MSST with the ability to compact the whole orientation information, multi-scale information description, and organic combination of frequency domain and spatial information, therefore, we adopt it to extract the multi-scale texture information. But, for MSST, it is a set of matrix that meets with the nature of Riemannian manifold, therefore, it will cost a great time consuming and memory burden when calculating the geodesic distance and statistic value by using space mapping from Riemannian manifold to tangent vector. More seriously, MMST is hard to integrate with color information directivity. To overcome these problems mentioned above, we use the SVD decomposition for each scale of MSST in tensor space, then, PCA algorithm used to extract the main texture information. Due to the fact that MSST does not work well in large-scale texture region, then, the total variation flow is used to auxiliary describe the texture feature by extracting local scale information. Meanwhile, to suppress the noises, enhance the boundaries and improve the homogeneous of similar color texture objects, the nonlinear diffusion filtering is adopted.Secondly, a component-wise expectation-maximization for MMST algorithm (CEM3ST) is proposed, it can calculate the valid class number through delete invalid component when using sample support degree of each component of mixed MMST distribution. To simulate and speed up the information propagation of valid class, we replace explicit scheme by semi-implicit scheme to update only one component at a time, and it can speed up the convergence. To exclude the impact of invalid component during the EM iterative procedure, the support degree of current component was calculated according to all the samples, using the treated and non-treated valid components. And then we can delete the invalid component after the sample support degree normalized. In other word, the invalid components can be quickly redistributed to the other valid components through recalculating the statistic value, valid class number, and sample support degree by adopting the least square and maximum likelihood (ML) process.Thirdly, we propose an edge-based and region-based multiphase successive active contour model (MSACM). In details, we improve the constraints about the external region force and internal boundary force of Chan and Vese (CV) model; it can enhance the homogeneous detection ability and noise tolerance through using the multivariable GMM model to describe probability density function (PDF). More important, we can break down the assumption about constant PDF in internal region. Except that, we incorporate geodesic active contour (GAC) into MSACM model to enhance the detection ability for concave edge and contrast with noise. As we knowledge, the optimal solution of MSACM model is equivalent to energy minimal, usually, we can resolve the optimal problem in numerical mode by using level set. Nevertheless, we confront with difficult, as the level set method is easily draped in local minimization and convergence slowly. As a result, we resort to Cauchy-Crofton formula; then, the energy minimization problem will be converted as maximum flow/minimum cut problem of multi-layer graph cut (MLG) model. So that we can quickly get the approximate optimal solution in global, using Graph Cut. It’s noteworthy that our proposed MSACM model, it cannot only captures the smooth and deep concave boundary, but also it can keep the segmented regions are more integrated finally.Lastly, an unsupervised color-texture image segmentation method is proposed using multivariate mixed student’s t-distribution (MMST) and regional credibility merging (RCM) strategy. To overcome the probability density function (PDF) appears the Gaussian, non-Gaussian, and heavy tailed distribution, we adopt the multivariate mixed student’s t-distribution to build up the PDF model. As it can integrate with CEM3ST algorithm for getting valid class number, thus, it can further reduce the layer number of MLG model greatly. Meanwhile, to overcome the segmented regions appear over-segmentation and error-segmentation phenomenon; a strategy of regional credibility merging (RCM) is presented by integrating the regional adjacency relationship, region size, common edge between regions, and regional color-texture dissimilarity-J divergence distance to calculate the RCM value for each two adjacent regions. If acquired RCM value is lower, the two adjacent regions will be merged or deleted. Additionally, in order to terminate the whole segmentation process adaptively, an adaptive iteration convergence criterion is designed, which combines the negative logarithm of probability of all color-texture features with the Kullback-Leibler (KL) divergence for MMST. So that the segmented regions with outperforming visual entirety and region consistency.