Dissertation > Excellent graduate degree dissertation topics show

Research on Panorama Stitching Based on Speeded Up Robust Features

Author: WeiYanLing
Tutor: QinHuaBiao；HuangRuoHao
School: South China University of Technology
Course: Electronics and Communication Engineering
Keywords: image mosaic image registration SURF image fusion multi-resolutionspline graph cuts
CLC: TP391.41
Type: Master's thesis
Year: 2011
Downloads: 90
Quote: 0
Read: Download Dissertation

Abstract


With the increasement of application requirements, image stitching technology hasbecome an important research subject in the field of virtual reality, computer vision, computergraphics, video processing and so on.This paper gives a systematic exposition of the panorama stitching technology based onits two key techniques: image registration and image fusion. Some improvements have beenproposed based on the current algorithms.In image registration, this paper elaborates the speeded up robust features (SURF)detection algorithm with the scale-invariant feature transform detection algorithm as acompare. In order to get a higher accuracy in the feature matching proceeding, this papermakes use of the feature pairs’symmetry property. The matching process is done by thefollowing steps: firstly, coarsely matched features are found by using the kd tree as a searchstructure and the nearest neighbor as a strategy; secondly, a vast majority of the falsematching pairs are discarded by judging whether they have symmetry or not; finally, atraditional random sampling verification (RANSAC) algorithm is implemented for a furtherfalse feature pairs discard and the image transformation parameters’calculation.In image fusion, this paper uses the multi-resolution spline technique combined withgraph cut-based seam search technique to achieve a smooth transition of the overlappingareas and ghost-free effects caused by moving objects in the input images. It also improvesthe initial weighted direction graph’s calculation which is based on the difference betweenpixel intensities. By considering not only the difference between pixel intensities but alsoimage gradient information, a more reasonable image cutting result can be achieved with theimproved weight calculation method. In addition, If the input images have a great exposuredifference, an unnatural color transition phenomenon will exist in the stitched image and theblock-based exposure compensation method can be implemented to fix it.Panorama stitching experiments have been done using algorithms described in this paper.The experimental data confirmes the feasibility and good result of the improved methods.

Related Dissertations

  1. Research of Image Mosaic Technology,TP391.41
  2. Study on Image Registration and Image Matching of the Robot System for Substation Equipment Inspection,TP242.62
  3. Study on the Image Registeration Integrated with the LQ Model,R815
  4. Information Extraction of Marine Oil Spill with Collaborative Images of ASAR and MODIS,X87
  5. The Research of Application of Ridgelet Transform in the Fusion of Multispectral and Panchromatic Images,TP391.41
  6. Multi-source image fusion technology research,TP391.41
  7. Research and Implementation of Non-rigid Medical Image Registration Method Based on Improved Optical Flow Model,TP391.41
  8. Study of Medical Image Registration Methods Based on Multi-features and Artificial Immune Algorithm,TP391.41
  9. Research of Medical Image Registration Method Based on DTCWT and NPSO,TP391.41
  10. Algorithm and Implementation Based on Non-rigid Medical Image Registration,TP391.41
  11. Integration of Spatial Information Bag of Feature in Image Annotation,TP391.41
  12. GPU-based SIFT algorithm,TP391.41
  13. Real-time video display system key technology research,TP391.41
  14. Object Detection and Accurate Localization in Industrial Application,TP391.41
  15. Research for Super-resolution Reconstruction Technologies of Image Based on POCS Algorithm,TP391.41
  16. Medical image registration platform and combines new gray with geometric information Registration Measures,TP391.41
  17. SURF algorithm based on multiple images of three-dimensional model reconstruction method,TP391.41
  18. SURF algorithm based on improved electronic image stabilization technology,TP391.41
  19. The mobile robot based on three-dimensional laser ranging Scene Recognition,TP242
  20. Graph-based color image segmentation algorithm targets,TP391.41
  21. SURF feature based on monocular vision SLAM Research and Implementation,TP242

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