Dissertation > Excellent graduate degree dissertation topics show

Blind Super-Resolution Image Reconstruct Based on Ggenetic Algorithm and Regularization

Author: WuWeiWei
Tutor: ZengQingShan
School: Zhengzhou University
Course: Control Theory and Control Engineering
Keywords: blind super-resolution restoration registration interpolation genetic algorithm regularization
CLC: TP391.41
Type: Master's thesis
Year: 2009
Downloads: 29
Quote: 0
Read: Download Dissertation

Abstract


In the process of capturing image, the original scene from the camera is easily warped because of the scene motion and the camera position change. In addition, it is difficult to obtain the clear high-resolution (HR) image due to various reasons, such as atmospheric turbulence, camera lens and the connatural limitation of digitized image sensor, etc. Base on the existing hardware conditions, how to acquire the HR from the low-resolution (LR) image would be the core of the Super Resolution (SR).The SR technology was developed nearly 50 years ago. The application of the early SR approach based on the signal frame image is limited because there is little helpful information could be used. The SR approach based on the multi-frame was proposed in 1980’s. It contains two approaches: frequency approach and space approach. As the classical SR model was founded, more and more attention has been devoted to study the space approach. In generally, it contains three elements: Registration, Interpolation and Remove the blur. Registration is the precondition of the SR technology. The accuracy of Registration is directly relation to the quality of the restoration image. The image interpolation and image restoration are integrated, which means to recover image on super-resolution grids. And there are more ill-conditioned and ill-posed than that of image restoration. The frame of regularization is introduced to restrict the ill-posed problem At the same, the genetic algorithm which is of better searching ability of global optimal solution is introduced to carry out regularization interpolation between the frames of image sequence.However, traditional technology of SR reconstruction assumes that the spread function of imaging system is a known point, and then handles the LR images for the HR image. How to achieve the blind super-resolution without any priori known blurs will be the new problem we faced. For overcoming these problems, the approach of blind super-resolution image reconstruct based on genetic algorithm is proposed; this approach introduces a regularization energy function, where regularization includes both the image and blurs domains. In order to enhance efficiency of the HR reconstruction and the quality of the image, we minimize the regularization by using global optimizing capacity of genetic algorithm. it achieve the reconstruction of blind super-resolution. In addition to that above, the rational down-sampling factor is introduced into the traditional integer down-sampling factor in this paper. The process of the rational down-sampling factor to be uses in the reconstruction of HR image is presented. The relation between the number of the LR images frames and SR factor is also illuminated. The results of the simulation is analyzed and compared with those of the traditional approaches. The simulation results demonstrate that the proposed algorithm is practical and applicable.

Related Dissertations

  1. The Study of a Radial Point Interpolation Meshless Method with Polynomial Basis and It’s Application,O241
  2. Development of the Platform for Compressor Optimization Design and Aerodynamic Optimization Design in the Transonic Compressor,TH45
  3. The Application of Fuzzy Comprehensive Evaluation Based on Genetic Algorithm in Vocational Evaluation of Classroom Teaching,G712
  4. BP network optimization based on genetic algorithm optimization of the biodiesel process,TE667
  5. Risk Assessment & Spatial Distribution of Soil Nutrients and Heavy Metal in Fenhe Reservoir Surrounding,S158
  6. Integer Polynomial factorization method,O174.14
  7. Research on FC+TCR Static Var Compensation Device,TM761.1
  8. China Regional Rainfall Changes of Space-time Mode Analysis,P426.614
  9. Research of Manifold Learning in the Problem of Sparse Sample Data,TP391.41
  10. Research and Implementation of Liver Visualization Based on Interpolation and Optimization,TP391.41
  11. Research on Multi-dimention and Multi-degree Dynamic Absorption,O328
  12. The Research and Realization of 2-PRR Parallel Machine Tool Control System,TG659
  13. The Application Research of Improved Harmony Search Algorithm in Vehicle Routing Problem,TP301.6
  14. Study in Mathematical Model on Heat and Moisture Transfer through Textiles,TS101
  15. Further Study on the Error Estimates for Least Squares Problems,O241.5
  16. On the Regularization Method Based Fuzzy C-Means Algorithm,TP311.13
  17. Electrical Impedance Tomography numerical solution of the problem in some studies,O441.4
  18. Study of numerical methods for Volterra integral equations of the first kind,O241.83
  19. On the BP Algorithm: A Regularization Approach,TP183
  20. Prediction Research of the Surrounding Rock Deformation in Engineering by the Support Vector Machine,TV223.1
  21. Research on Image Restoration with Combined Bases,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