Dissertation > Excellent graduate degree dissertation topics show

Research on Image Segmentation and Scene Understanding Using Discriminative Learning Method

Author: WangJianLe
Tutor: HuangYaPing
School: Beijing Jiaotong University
Course: Computer Science and Technology
Keywords: Texture features Geometric characteristics Discriminant Texture image segmentation Scene Image Understanding
CLC: TP391.41
Type: Master's thesis
Year: 2011
Downloads: 103
Quote: 0
Read: Download Dissertation

Abstract


Image segmentation is a key and difficult issues of the computer vision field , this paper, a new fusion of texture and geometric characteristics of the framework of image segmentation , image segmentation problem into a classification problem to be solved . Handled by different objects , further study of the problem of the choice of features and discriminant learning method , and in texture image segmentation , outdoor scenes and indoor scene images were discussed in depth in the understanding of three typical applications . The main work of the paper are: a fusion of texture and geometric features discriminant framework for image segmentation , optimization target segmentation problem into a classification problem . For texture image segmentation , further divided into two sub-problems interactive segmentation based on texture features and texton - based scene segmentation study . Gradient histogram features and boosting algorithm for interactive segmentation based on texture features used to split processing ; based on texton scene segmentation using texton feature (textons) and Joint Boosting algorithm to split . Experimental results show that the texture feature - based discriminant learning method , can effectively distinguish the texture pattern in image recognition and segmentation and image blocks in various scenes images of different semantic categories . 3 , on the basis of the above study , further discussion of the three-dimensional scene to understand the problem . Image for the outdoor scenes and indoor scene images , the proposed algorithm based on the geometric characteristics of image segmentation and scene understanding . The experimental results show that even in a variety of complex scenes , the method was still able to get a better segmentation results .

Related Dissertations

  1. Analysis on Characteristics and Discriminant of Aroma Types of Flue-Cured Tobacco,S572
  2. The Research and Implementation of Lgorithms on Microcalcification Detection,R816.4
  3. Face recognition algorithm based on feature fusion research,TP391.41
  4. Data Driven Rolling Bearing Fault Diagnosis Research,TH165.3
  5. Keyframe-based video object segmentation hybrid modeling technology research,TP391.41
  6. Single Building in the rapid assessment study,TU311.3
  7. Facial feature extraction system is a key technology research and its DSP,TP391.41
  8. Under the conditions of variable lighting and occlusion of face recognition technology and its application,TP391.41
  9. PTT \,TS101.923
  10. Credit risk management techniques and profit forecasting model,F224
  11. Regional innovation system in Zhejiang Province,F127
  12. Discriminative Sparse Coding Methods for Human Action Recognition in Video,TP391.41
  13. Preliminary Metabolomics Study of Breast Cancer Based UPLC-Q/TOF and ~1 HNMR,R737.9
  14. Serum Metabolite Profiling of the Hepatitis B Virus Related Cirrhosis,R512.62
  15. Application of Partial Least Square and Discrimnent Analysis in Studying the Style and Influence Factors of the Scientific Personnel,G644
  16. Research on Region Target Semi-automatic Extraction of RS Images Based on Texture Features,TP751
  17. Research on Region Target Semi-automatic Extraction of RS Images Based on Texture Features,TP751
  18. Research on Change Detection for Remote Sensing Images Based on Discriminant Analysis,TP751
  19. The Quantitative Retrospective Study in Grade Level of Brain Dysfunction Induced by Head Injury (Ⅰ),D919
  20. Screening and Value of Oxygenic Injury Indexes of Coal Workers’ Pneumoconiosis,R135.2
  21. The Study on Modern Processing Method and UPLC-DAD-TOF-MS Fingerprint of Wen-Hou-Po,R284

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