Dissertation > Excellent graduate degree dissertation topics show

Adaptive Markov random weights Color texture image segmentation

Author: LiuZuo
Tutor: HuangXueYu
School: Jiangxi University of Technology
Course: Applied Computer Technology
Keywords: MRF Gibbs color texture picture adapting weight MRF minimum energy arithmetic
CLC: TP391.41
Type: Master's thesis
Year: 2008
Downloads: 173
Quote: 0
Read: Download Dissertation

Abstract


Human eyes are more sensitive to colors than intensities and there exists more information in color images which can bring richer perception. For a long period, most researches in vision fields have focused on gray-level images. In recent years, the increasing performance of computer hardware and image capturing equipments and their decreased cost make color image processing possible. Furthermore, with the rapid development of multimedia techniques, more and more color images need to be processed(such as print images),and color image processing, especially color image segmentation, becomes an important topic in image processing area gradually. As color image devices are becoming more and more popular nowadays, it is really stringent to accelerate color image processing techniques. From two aspects of theoretical study and practical application, this paper pays more attention on several key issues in color image processing, i.e. color image feature extraction and color image segmentation. Image Segmentation is a process in which the image is segmented into different homogeneous regions. In other words, finding the edges among these regions can achieve this goal. Comparing with gray image, color image contains not only intensity information, but also other useful information such as tonality, saturation. In recent years, image retrieval based on image content, color and texture has been a focus in database technique, in which color image segmentation is the basic technique.Markov Random Field has been widely applied to solve the problem fields of image processing and machine vision. MRF provides a straight way, which is based on decision- theory, to model the relationship between pixels. An equivalence relation between MRF and Gibbs distribution make MRF model be widely applied, and the joint distribution provides a model which researchers can used with Bayesian to solve the problems coming from image processing and machine vision.We propose an adapting weight MRF image segmentation model, which aims at combining color and texture features. The theoretical framework relies on Bayesian estimation via combinatorial optimization (simulated annealing).The segmentation is obtained by classifying the pixels into different pixel classes. These classes are represented by multi-variate Gaussian distributions. Thus, the only hypotheses is about the nature of the features is that an additive Gaussian noise model is suitable to describe the feature distribution belonging to a given class. Here, we use the perceptually uniform HSV color values as color features and a set of Gabor filters as texture features. Gaussian parameters are either computed using a training data set or estimated from the input image. We also propose a parameter estimation method using the EM algorithm.There are three advantages for our model; firstly, the time used for computer is reduced for we just use a simple single layer MRF model. Secondly, we use a combination of classical, gray-level based, texture features and color instead of direct modeling of color textures. Hence, most of the classical texture features can be used. Thirdly, we use a new adapting weight MRF model, which make the results better and artificial intervention reduced.In chapter 5, we designed 4 experiments to illustrate the performance of our method on both synthetic and natural color image.

Related Dissertations

  1. Bayesian methods under two ordinal value of multi-value data model with simultaneous identification of outliers,O212.1
  2. Exam Quantity Analysis and Student’s Ability Assessment Based on IRT Model,G449
  3. Coupled silicon and silicon tetrachloride hydrogenation reaction of trichlorosilane Process,O613.72
  4. Using frequency domain analysis of the details of the method of generating fluid animation,TP391.41
  5. Mold cavity surface MRF technology research,TG580.692
  6. Bayesian Methods for Gross Errors Detection of Correlated Observations and Its Application in GPS Network Adjustment,P228.4
  7. Financial Stochastic Volatility Extension Model Analysis and Applied Research,F830
  8. Obstacle Detection Method Based on Single Camera,TP242
  9. Algorithm for Gibbs Sampling Motif Discovery Based on Position Interdependency,TP301.6
  10. Research and Application on User-Controllable Multi-Exemplars Texture Synthesis,TP391.41
  11. Duration Dependence of the Business Cycle in China,F124.8
  12. An Effective-field Theoretical Study on Ising Metamagnet,TB34
  13. The Study of Stereo Matching Using Belief Propagation,TP391.41
  14. Classification of the Seafloor Based on the Sonar Image Analysis,TP391.41
  15. Research on Change Detection Technology of SAR Image,TN957.52
  16. Research on Segmentation Methods Based on MRF Model from SAR Images,TN957.52
  17. Algorithm Study of SPA Based on Mixed-resolution Data,F425
  18. Gibbs sampling inference algorithm based traffic forecast study,TP18
  19. Research and Implementation of MRF Exemplar-based Inpainting Method in Wavelet-domain,TP391.41
  20. Research of Financial Time Series Based on Markov Switching Cointegration Model,F830.9
  21. Mathematical Calculation of the Biomass Gasification,TK6

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net  Mobile