Dissertation > Excellent graduate degree dissertation topics show

Analysis and Comparison of Image Salient Region Extraction Algorithms Based on Attention Mechanism

Author: LiMinXue
Tutor: SongZeHai
School: Beijing Jiaotong University
Course: Computer Science and Technology
Keywords: Visual perception Saliency map Significant regional Algorithm evaluation
CLC: TP391.41
Type: Master's thesis
Year: 2011
Downloads: 223
Quote: 2
Read: Download Dissertation

Abstract


Image based on biological perception of the salient features of the analysis of the rise in the late 1990s, and biological visual perception becoming the focus of research in the field. The method combines the knowledge of human psychology and physiology theory, establish the image significantly extraction model based on human visual attention mechanism to simulate the human eye function. Visual feature extraction significantly as a separate technology, can help us to better analyze and understand digital images. Extract images sex research is an image analysis, feature extraction and comprehensive process to explore the characteristics of human visual-based image analysis and understanding applications have very important significance. The work of this paper is to analyze the various saliency map generated model-based attention mechanism significantly graph generation algorithm developed algorithm and compare systems, and the results of analysis. This paper analyzes the five kinds of saliency map generation algorithm. Itti model is typical of the attention to the space-based model, it has an input image extract color, brightness, direction and other characteristics, and then formed in the dimension of each feature attention to Figure, the ultimate fusion of these concerns Pictured saliency map; SMG (Saliency map generation) algorithm is a \) algorithm is based on regional contrast the significantly digraph generation algorithm, the algorithm using segmentation algorithm to image segmentation into regions based on the color histogram of the significant forward value calculation on the regional level; the FT (the Frequency-Tuned) algorithm is based on the frequency FM of significantly with map generating algorithm, and to increase the edge detection. Based on the research work, I extracted significant feature-based attention mechanism theory of the sort, to re-implement the Itti algorithms and SMG algorithm. Significantly the accuracy of the region and then according to the extraction of the algorithm, the comparative evaluation of each algorithm, reference standard Microsoft Research Asia (MSRA) provides significant object image library. Comparison, the RC algorithm accurately rate the highest, but the storage capacity of the algorithm and a large amount of computation: SMG algorithm efficiency high system overhead, to help meet the requirements of real-time image retrieval; Itti algorithm needs to drop the original image sampling, significant only 1/256 of the original image; Figure FT algorithm to calculate high efficiency and output saliency map full resolution; the LC algorithm avoids image segmentation, and therefore difficult to automatically split the image, the effect is very good.

Related Dissertations

  1. Visual Feedback and Memory Behavior Based GPU Parallel Ant Colony Algorithm,TP301.6
  2. Segmentation of cDNA Microarray Image Using Fuzzy C-means Algorithm Optimized by Particle Swarm,TP391.41
  3. A Method for Image Quality Assessment Based on Visual Attention and Structure Distortion,TP391.41
  4. The Research of Applications and Industrial Design in Police UAV,V279
  5. Saliency -based image retrieval keying Research and Implementation,TP391.41
  6. Temporal integration of visual selective attention model and applied research in the MPEG compressed domain,TP391.41
  7. Low-quality document image binarization research,TP391.41
  8. Bayesian framework of image saliency detection,TP391.41
  9. Research and Implementation of Computer Game Strategy Based on Reinforcement Learning,TP18
  10. Based on the topological properties of the visual attention model and its application,TP391.41
  11. Research on User-Interface Design of Multimedia Electronic Atlas,P208
  12. Art Design and Modelling Ability’s,J506
  13. The Intrepretation of the Creative Thinking with Semantics of Visual Forms,J524
  14. Cryptographic Algorithm Automatic Evaluation System,TP309
  15. A Study of Theory and Technology of Planar Animated Map,P208
  16. The Research on the Space Orgnization of Lingering Garden Based on the Visual Sensation,TU986.1
  17. Study on the VI Design of Mental Hospital Base on Visual Perception,TU246.1
  18. High-performance Remote-Sensing Image Fusion Algorithm Research and It’s Implementation,TP751
  19. Urban Landscape Logo Color Analysis and Research,J524
  20. A New Color Image Segmentation Algorithm Based on Region Growing,TP391.41
  21. Recognition and Coding of ROI Based on Intelligence Surveillance,TP391.41

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