Dissertation > Excellent graduate degree dissertation topics show

Research on Constructing A Lexica Family of Concepts with Small Semantic Gap for Image Retrieval

Author: LiuJieMin
Tutor: YangXiaoKang；ShiPeng
School: Shanghai Jiaotong University
Course: Communication and Information System
Keywords: Image Retrieval Semantic gap Shallow semantic divide thesaurus Affinity Propagation Clustering
CLC: TP391.3
Type: Master's thesis
Year: 2010
Downloads: 99
Quote: 1
Read: Download Dissertation

Abstract


In recent years, image retrieval become important research topic in the field of multimedia information retrieval. \Semantic-based image retrieval technology, the establishment of effective object recognition or automatic annotation semantic concept model in order to shorten the semantic gap. A good definition of semantic concepts the library these methods in data collection, the first step in modeling is a crucial step. The semantic gap semantic concepts inherent vary, information processing with image understanding method is also far less than the extracted image abstract (deep) semantic requirements. More realistic approach is to try to find those computers that are easy to learn the the shallow semantic gap semantic concept, these semantic concept more useful concept detection model of training, followed by semantic recognition and automatic annotation. Therefore, to identify the the shallow semantic divide thesaurus for semantic-based image retrieval technology has important significance, involves two main issues: 1) how to define shallow semantic gap \? 2) how to automatically identify such semantics? the work done by the subject is an innovative way to solve these two problems, and ultimately build a shallow semantic gap thesaurus the semantic thesaurus in the study of large-scale image retrieval of data collection Feature selection, build retrieval model, image annotation provides useful suggestions. This paper first elaborated the to build the shallow semantic divide thesaurus basic framework: 1) 2.4 million Internet image extract semantic text feature, as well as a variety of low-level visual features, namely the establishment of effective index. 2) in different semantic gap model, calculated every picture visual - text confidence, is a measure of the the two distributions consistency of the image and its close neighbors in the visual feature space and text feature space. 3) the use of affinity propagation clustering algorithm to cluster the image with the highest visual - text confidence. 4) from the clustering results of keyword extraction based on text content most relevant keywords is the most shallow semantic concepts of the semantic gap. Different visual space semantic gap, starting from the number of low-level visual space, respectively, based on color feature, texture feature and color texture characteristics, to build the shallow semantic gap vocabulary. Comparative analysis of their similarities and differences, been based on the the shallow semantic gap lexicon of visual features, can provide effective advice semantic concepts in image retrieval feature selection. In this paper, two duality the semantic gap model - text diffusion model and visual diffusion model for the distribution of the image in the visual space and text space inconsistencies. In essence, the two semantic gap model correspond to the visual content-based retrieval and text-based content retrieval. Integrated by two models the shallow semantic gap thesaurus can choose the search for semantic concept can be applied to the optimization of image annotation. Affinity propagation clustering algorithm to solve large-scale image clustering problems. The clustering algorithm has four major advantages: 1) determine the number of category clustering without. 2) the input requirements of the similarity matrix. The need to consider both the visual and text twofold image clustering similarity, the similarity matrix is ??more reasonable and effective than the use of high-dimensional data points. 3) also apply to the asymmetry of similarity between the two images. 4) can effectively handle large data sets. A large number of experimental data fully: paper constructs the shallow semantic gap thesaurus the shallow semantic gap vocabulary independently of each other to complement each other, the data collection in the study of large-scale image retrieval, the low-level feature selection, efficient retrieval mode selection and image labeling and other sectors played an important role, and provides a new way of thinking is based on the development of semantic image retrieval technology.

Related Dissertations

  1. Research on Facial Feature Extraction and Matching Algorithms for Image Retrieval,TP391.41
  2. The Research and Implemention of Image Retrieval Based on User Interested Feature,TP391.41
  3. Application of Q-Learning in the Content-Based Image Retrieval Technology,TP391.41
  4. Research and Implementation on Content-Based Clothing Image Retrieval,TP391.41
  5. The Discovery of User Concept Region Based on Multiple Instance Learning,TP391.41
  6. Research on Transductive Support Vector Machine and Its Application in Image Retrieval,TP391.41
  7. Research and Application of Diverse Density Learning Algorithm,TP181
  8. Jade Multi-Agent Based Image Retrieval System,TP391.3
  9. Multi- mode image retrieval method,TP391.41
  10. Research on Shape Based Algorithm for Image Retrieval,TP391.41
  11. Image Retrieval Based on SIFT Feature Point Detection,TP391.41
  12. Based on independent component analysis for image retrieval research,TP391.41
  13. Research on Automatic Image Annotation,TP391.41
  14. Research in Image Retrieval Based on Integrated Features and Relevant Feedback,TP391.41
  15. Content-based color image retrieval,TP391.41
  16. Research on the Techniques for Color-Based Image Retrieval,TP391.41
  17. Shape Feature Extraction and Retrieval Technology,TP391.41
  18. Integration into high-level semantic feature of medical image retrieval technology,TP391.3
  19. Features and relevance feedback based on the underlying medical image retrieval,TP391.41
  20. PCA image classification based on the extended model,TP391.41
  21. Based on the Edge-Characteristics of Image Content Retrieval Research,TP391.41

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Retrieval machine
© 2012 www.DissertationTopic.Net  Mobile