Figure 1 - uploaded by Joachim M Buhmann
Content may be subject to copyright.
Source publication
Unsupervised image segmentation can be formulated as a clustering problem in which pixels or small image patches are grouped together based on local feature information. In this contribution, parametric distributional clustering (PDC) is presented as a novel approach to image segmentation based on color and texture clues. The objective function of...
Context in source publication
Context 1
... therefore, represent individual groups by continuous mix- ture models, which are adapted by parameter fitting. This modeling decision has the advantage that, when matching a locally measured histogram to those continuous models, the ordering of the bins is no longer disregarded ( fig.1). ...
Similar publications
For the problem that the different parameters of infrared imaging equipment and the environment around the target cause the poor robustness of threshold value automatic acquisition method in infrared human target segmentation algorithm, starting from the principle of infrared imagery and connecting with the characteristics of the histogram and K-me...
Citations
... Brox et al. [58] propose to combine colour, texture and motion in a level-set image segmentation framework. Zoller et al. [59] formulated the image segmentation as a data clustering process. A related image segmentation approach was proposed by Ooi and Lim [60]. ...
This literature review attempts to provide a brief overview of some of the most common segmentation techniques, and a comparison between them.It discusses Graph based methods, Medical image segmentation research papers and Color Image based Segmentation Techniques. With the growing research on image segmentation, it has become important to categorise the research outcomes and provide readers with an overview of the existing segmentation techniques in each category. In this paper, different image segmentation techniques starting from graph based approach to color image segmentation and medical image segmentation, which covers the application of both techniques, are reviewed.Information about open source software packages for image segmentation and standard databases are provided. Finally, summaries and review of research work for image segmentation techniques along with quantitative comparisons for assessing the segmentation results with different parameters are represented in tabular format, which are the extracts of many research papers. Index Terms—Graph based segmentation technique, medical image segmentation, color image segmentation, watershed (WS) method, F-measure, computerized tomography (CT) images I. Introduction Image segmentation is the process of separating or grouping an image into different parts. There are currently many different ways of performing image segmentation, ranging from the simple thresholding method to advanced color image segmentation methods. These parts normally correspond to something that humans can easily separate and view as individual objects. Computers have no means of intelligently recognizing objects, and so many different methods have been developed in order to segment images. The segmentation process in based on various features found in the image. This might be color information, boundaries or segment of an image. The aim of image segmentation is the domain-independent partition of the image into a set of regions, which are visually distinct and uniform with respect to some property, such as grey level, texture or colour. Segmentation can be considered the first step and key issue in object recognition, scene understanding and image understanding. Its application area varies from industrial quality control to medicine, robot navigation, geophysical exploration, military applications, etc. In all these areas, the quality of the final results depends largely on the quality of the segmentation. In this review paper we will discuss on graph based segmentation techniques, color image segmentation techniques and medical image segmentation, which is the real time application and very important field of research. The mathematical details are avoided for simplicity.
... Zoller et al [78] formulated the image segmentation as a data clustering process where the colour and texture features were combined using a parametric distributional clustering (PDC) approach. In this implementation the colour features were given by the three components of the HSV colour space, while the texture features were extracted by convolving the V component of the image with a set of Gabor filters with four orientations. ...
The adaptive integration of the colour and texture attributes in the development of complex image descriptors is one of the most investigated topics of research in computer vision. The substantial interest shown by the research community in colour–texture-based segmentation is mainly motivated by two factors. The first is related to the observation that the imaged objects are often described at perceptual level by distinctive colour and texture characteristics, while the second is motivated by the large spectrum of possible applications that can be addressed by the colour–texture integration in the segmentation process. Over the past three decades a substantial number of techniques in the field of
colour–texture segmentation have been reported and it is the aim of this article to thoroughly evaluate and categorise the most relevant algorithms with respect to the modality behind the integration of these two fundamental image attributes. In this paper we also provide a detailed discussion about data collections, evaluation metrics and we review the performance attained by state of the art implementations. We conclude with a discussion that samples our views on the field of colour–texture image segmentation and this is complemented with an examination of the potential future directions of research.
... Medidas de similaridade para informações sobre cor e textura, muito importantes em técnicas que buscam identificar regiões defeituosas com base em amostras de couro sem e com defeitos, foram avaliadas empiricamente em [Puzicha and Bhumann 1999]. [Zoller et al. 2002] também apresentam uma experiencia interessante de combinação de informações relacionadas com cor e textura para segmentação de imagens complexas. ...
A classificaçclassificaç˜classificação, com base na inspeçinspeç˜inspeção visual e detecçdetecç˜detecção de proble-mas como marcas de fogo, riscos e placas de berne, entre outros, ´ e uma fase importante no processamento do couro bovino. A automatizaçautomatizaç˜automatização desse pro-cesso, utilizando técnicas de processamento digital de imagens, pode torná-lo mais preciso, menos subjetivo, menos sujeito a falhas e mais uniforme. Este pro-jeto tem como objetivo a construçconstruç˜construção de um detector de imperfeiçimperfeiç˜imperfeições no couro utilizando técnicas de visão computacional baseadas em análise de textura. 1. IntroduçIntroduç˜Introdução Apesar de possuir o maior rebanho bovino do planeta, o Brasil ainda não conseguiu as-sumir um papel de destaque na indústria de couros. A maior parte de couro produzido no paíspaís´paísé de baixa qualidade, sendo comercializados geralmente, nos estágios iniciais de pro-cessamento, com baixo valor agregado. A baixa qualidade do couro resulta, em grande parte, de problemas que poderiam ser facilmente solucionados, como o uso de ferrão, cercas de arame farpado, marcaçmarcaç˜marcação a ferro quente em local impróprioimpr´impróprio, entre outros. A quase inexistência de sistemas de remuneraçremuneraç˜remuneração que valorizem a qualidade do couro, ´ e ge-ralmente citado como o principal inibidor de investimentos no aumento da qualidade do couro bovino.
... This approach, tied together with metadata on images such as histograms, texture, and color sampling, has the potential to yield high-precision results in image classification. Examining the textual descriptions associated with an image provides additional information that may be used to classify the image more accurately [24, 45]. Unfortunately, this approach does not take into account the connections among individual objects presented in a sample image . ...
The key problem in achieving efficient and user-friendly retrieval in the domain of image is the development of a search mechanism
to guarantee delivery of minimal irrelevant information (high precision) while ensuring that relevant information is not overlooked
(high recall). The unstructured format of images tends to resist the deployment of standard search mechanism and classification
techniques. As a method to provide better organization of images, clustering is an important aspect for effective image retrieval.
... Boolean assignment variables M iν ,ν = 1, ..., k encode the cluster memberships of a site s i ; so we actually set M iν = 1, if site s i is assigned to cluster ν. We further impose the constraint that i≤ν M iν = 1 holds to have a well defined cluster assignment process and to avoid multiple cluster assignments per site [3]. With p ν we denote the probability of cluster ν with respect to the actual image data, with ν the unique and characteristic mode -based on the histogram of the noise removed image data Y of course-of cluster ν also with respect to the actual data and with µ, σ 2 the mean and variance of an identical independent Gaussian distribution G µ,σ 2 (y i ) representing our noise model. ...
In this contribution we describe a Bayesian modeling frame-work for low-level image-processing problems which (1) formulates the problems under consideration on a regular Markovian pixel grid and (2) uses distributions with Gibb-sian form for the modeling which have maximum entropy among the family of distributions. This modeling frame-work possesses the general properties of massively parallel and spatially local image-entity calculations and serial and global model-parameter estimation. For this class of image-processing models we propose a massively parallel digital grid-architecture. The proposed modeling approach and the corresponding grid-architecture is extensively tested on sev-eral images by means of a noise removing and unsupervised segmentation model. Chip design examples of the building blocks in FPGA-technologies are also included to finalize this contribution.
... Image segmentation methods can be further classified according to the used image cues. Several methods integrate different image cues such as texture and color or brightness (Belongie et al., 1998;Malik et al., 1999;Manduchi, 1999;Thirion et al., 2000;Panjwani and Healey, 1995;Zhong et al., 2000;Zeoller et al., 2002;Mirmehdi and Petrou, 2000). We use local statistics which jointly characterize texture, color, and brightness. ...
The task of tting parametric curve models to the boundaries of perceptually meaningful image regions is a key problem in computer vision with numerous applications, such as image segmentation, pose estimation, object tracking, and 3-D reconstruction. In this article, we propose the Contracting Curve Density (CCD) algorithm as a solution to the curve-tting problem. The CCD algorithm extends the state-of-the-art in two important ways. First, it applies a novel likelihood function for the assessment of a t between the curve model and the image data. This likelihood function can cope with highly inhomogeneous image regions, because it is formulated in terms of local image statistics. The local image statistics are learned on the y from the vicinity of the expected curve. They provide therefore locally adapted criteria for separating the adjacent image regions. These local criteria replace often used predened x ed criteria that rely on homogeneous image regions or specic edge properties. The second contribution is the use of blurred curve models as efcient means for iteratively optimizing the posterior density over possible model parameters. These blurred curve models enable the algorithm to trade-off two conicting objectives, namely heaving a large area of convergence and achieving high accuracy. We apply the CCD algorithm to several challenging image segmentation and 3-D pose estimation problems. Our experiments with RGB images show that the CCD algorithm achieves a high level of robustness and sub- pixel accuracy even in the presence of severe texture, shading, clutter, partial occlusion, and strong changes of illumination.
... For an in-depth presentation and derivation we have to refer the interested reader to additional literature [3,4,2]. ...
... Some results of this particular simulation series are depicted inFig- ure 3. These results are comparable with state-ofthe-art results of [2]. To further underpin and prove the practical respectively industrial relevance ofFig. ...
In this paper we propose the structure of a novel massively parallel system-on-chip (SoC) architecture for a state-of-the-art probabilistic im-age segmentation model. This probabilistic model is formulated on a regular Markovian pixel grid. The unique combination of algorithmic robustness, SoC and real-time processing capabilities provides a new type of image processing systems in real-world applications. The model and the SoC archi-tecture are extensively tested by natural images. Some chip design examples are also included to fi-nalize the contribution.
... Image segmentation methods can be further classified according to the used image cues. Several methods integrate different image cues such as texture and color or brightness (Belongie et al., 1998;Malik et al., 1999;Manduchi, 1999;Thirion et al., 2000;Panjwani and Healey, 1995;Zhong et al., 2000;Zeoller et al., 2002;Mirmehdi and Petrou, 2000). We use local statistics which jointly characterize texture, color, and brightness. ...
In this article we propose the Contracting Curve Density (CCD) algorithm, a novel algorithm for fitting parametric curve models to image data. The CCD algorithm contributes to the area of model-based image segmentation in two important ways. First, for the assessment of a fit between the curve model and the image data, the CCD algorithm employs a likelihood function that is capable of coping with highly inhomogeneous image regions. This capability is achieved by formulating the likelihood function in terms of local image statistics that are iteratively learned from the vicinity of the expected curve. The local statistics provide locally adapted criteria for separating adjacent image regions. This replaces often used global predefined criteria, relying on homogeneous image regions or specific edge properties, such as a particular edge profile or an image gradient exceeding a preset threshold. Second, for the optimization of the resulting posterior density, we propose an iterative approach based on a blurred curve model. Blurred curve models enable the algorithm to trade-off two conflicting objectives: large area of convergence and high accuracy. We apply the CCD algorithm to several image segmentation and 3-D pose estimation problems. Our experiments with RGB images show that the CCD algorithm is capable of achieving high subpixel accuracy and robustness even in the presence of severe texture, shading, clutter, partial occlusion, and strong changes of illumination.
A novel level set based image segmentation method is proposed in this paper. After analyzing advantages and drawbacks of SBGFRLS and LSD model, we propose a way to combine local information and global information through utilizing weighted energy generated by SBGFRLS model to reduce segmentation error and accelerate curve evolution. The total energy of proposed method comprises of local energy term, length term and weighted global energy term. In our experiments, two alternative values for the coefficient of global term are 0.1 or -0.1. Experiments on images with noise and intensity inhomogeneities show that the proposed method is effective and more accurate than both SBGFRLS model and LSD model.
