In this paper we examine the problem of image segmentation of noisy images. We consider a doubly stochasticimage model. The image is assumed to be the sum of the realizations of two independent random fields: the unc...
详细信息
In this paper we examine the problem of image segmentation of noisy images. We consider a doubly stochasticimage model. The image is assumed to be the sum of the realizations of two independent random fields: the uncorrupted image and the noise field, consisting of independent, identically distributed, Gaussian random variables. The image segmentation technique employed here is a technique in which the image is represented by a semi-Markov random field corrupted by additive white noise. An adaptive Bayesian parameter estimation/image detection algorithm is developed. This algorithm allows us to estimate the unknown image and its underlying parameters in an optimal manner. We demonstrate the potential of the proposed algorithm in the case of the smoothing/segmentation of two 4-gray level real images.
Edge detection in sampled images may be viewed as a problem of numerical differentiation. In fact, most point edge operators function by estimating the local gradient or Laplacian. Adopting this view, Torre and Poggio...
详细信息
Edge detection in sampled images may be viewed as a problem of numerical differentiation. In fact, most point edge operators function by estimating the local gradient or Laplacian. Adopting this view, Torre and Poggio [2] apply regularization techniques to the problem of computing derivatives, and arrive at a class of simple linear estimators involving derivatives of a low-pass Gaussian kernel. In this work, we further develop the approach by examining statistical properties of such estimators, and investigate the effectiveness of various combinations of the partial derivative estimates in detecting blurred steps and lines. We also touch briefly on the problem of sensitivity to various types of edge structures, and develop an isotropic operator with reduced sensitivity to isolated spikes.
The solution of many estimation problems can be greatly enhanced by the incorporation of inexact knowledge or vague human reasoning. For such estimation problems, two distinct forms of problem knowledge can be identif...
详细信息
The solution of many estimation problems can be greatly enhanced by the incorporation of inexact knowledge or vague human reasoning. For such estimation problems, two distinct forms of problem knowledge can be identified: statistical (objective) knowledge and heuristic (subjective) knowledge. This paper discusses a systematic way of expressing and integrating these two forms of knowledge into the estimation process. This work can be interpreted as a fuzzification of standard constrained optimization. Fuzzy set theory is used to form a fuzzy constraint which represents the domain-specific knowledge of human expert. This work may also be interpreted as a systemization of the use of subjective priors by Bayesians. Although our work is of general applicability, we demonstrate the use of heuristically constrained estimation to the particular problem of seismic deconvolution. These results show that the incorporation of heuristic knowledge (albeit vague) yields better results than if such knowledge is ignored.
Discrete median filters are a special class of ranked-order digital filters used for smoothing signals. In this paper, the analog median filter is defined and proposed for analysis of the standard discrete median filt...
详细信息
Discrete median filters are a special class of ranked-order digital filters used for smoothing signals. In this paper, the analog median filter is defined and proposed for analysis of the standard discrete median filter in cases with a large sample size or when the associated statistics would be simpler in the continuum. Discrete filters are shown to be a subclass of analog filters. Also, an equivalence among analog filters and limits of discrete filters is established. Finally, several stochastic interpretations of the analog median filter are presented including necessary and sufficient conditions on input processes which guarantee the existence of output distributions for multiple passes of the analog median filter
stochastic sampling is a good alternative to pure oversampling in terms of image quality. A method for adaptively controlling the number of required samples to the complexity of the picture is presented. The quality o...
详细信息
ISBN:
(纸本)044470065X
stochastic sampling is a good alternative to pure oversampling in terms of image quality. A method for adaptively controlling the number of required samples to the complexity of the picture is presented. The quality of the obtained picture can be controlled by two well-understandable parameters, these parameters define an error interval size and the probability that a pixel lies within it. The usefulness of the method is described by applying it to distributed ray-tracing.
The main objectives of this paper are to relate some of the relaxation labelling algorithms to methods in the literature on the statistical analysis of incomplete data and to investigate to what extent this helps to r...
详细信息
ISBN:
(纸本)0852963325
The main objectives of this paper are to relate some of the relaxation labelling algorithms to methods in the literature on the statistical analysis of incomplete data and to investigate to what extent this helps to render more systematic the relaxation procedures.
In this study, we consider an image restoration problem in which the impulse response varies randomly with time (due e. g. to random fluctuations of the system's pupil function) while the object f(x) remains fixed...
详细信息
ISBN:
(纸本)0444700684
In this study, we consider an image restoration problem in which the impulse response varies randomly with time (due e. g. to random fluctuations of the system's pupil function) while the object f(x) remains fixed. The recorded image is assumed to be measured in the time interval left bracket O, T right bracket . f is to be estimated assuming that some statistical properties of the stochastic process are known.
One of the most successful second-order statistical representations for texture is the gray-level cooccurrence matrix. Many scalar features from this matrix have been suggested. An approach to texture classification a...
详细信息
ISBN:
(纸本)0818607424
One of the most successful second-order statistical representations for texture is the gray-level cooccurrence matrix. Many scalar features from this matrix have been suggested. An approach to texture classification and clustering is described that is based on two objectives: first, the information loss due to feature extraction is alleviated by using the cooccurrence matrix directly as the feature vector: second, the classifier is adapted to the feature space by using the subspace method. Results on subspace clustering and classification of textures, using some natural and some synthetic examples, compare favorably with other texture feature extraction and classification methods.
Texture provides an important source of information about the local orientation of visible surfaces. To recover 3-D structure, the distorting effects of the perspective projection must be distinguished from properties...
详细信息
ISBN:
(纸本)0818607211
Texture provides an important source of information about the local orientation of visible surfaces. To recover 3-D structure, the distorting effects of the perspective projection must be distinguished from properties of the texture on which the distortion acts. This requires that assumptions must be made about the texture. methods are reported that exploit the uniform density assumption, first introduced by J. J. Gibson (1950), as well as a generalization of Gibson's assumption that involves the sum of the lengths of the edges per unit area. The conditions under which either the number of texels can be counted or the boundaries of the texels can be located, are discussed. The strategy has been implemented using statisticalmethods for the case of textured planes. The algorithms, which are conceptually simple and computationally effective, have been tested with both synthetic and natural images and the results were very satisfactory.
In this paper, the Toeplitz Approximation Method (TAM) of stochastic system identification is applied to the linear equal spaced array narrowband source direction finding problem. The proposed algorithm provides high ...
详细信息
In this paper, the Toeplitz Approximation Method (TAM) of stochastic system identification is applied to the linear equal spaced array narrowband source direction finding problem. The proposed algorithm provides high resolution direction finding capability and is designed for an arbitrary noise, multipath signal environment. As such, it extend existing capability in fields such as passive sonar, radar and communications. A comparitive simulation between TAM and the MUSIC method, using spatial smoothing, is presented which are based on low signal-noise-ratio (SNR) data and a multipath environment.
暂无评论