Dissertation > Excellent graduate degree dissertation topics show

Research of image inpainting algorithm based on TV model and evolution

Author: YangZhen
Tutor: LiKangShun
School: Jiangxi University of Technology
Course: Applied Computer Technology
Keywords: Image completion Total variation Evolutionary algorithm Texture synthesis
CLC: TP391.41
Type: Master's thesis
Year: 2013
Downloads: 67
Quote: 0
Read: Download Dissertation

Abstract


With the development and improvement of electronic products and communicationtechnologies, more and more digital images appear in people’s daily life. While in reality, manyman-made or objective factors will cause images damaged. For a simple and effective way torecover the damaged images that have a certain value and significance, more and moreresearchers constantly study in digital image completion techniques. In recent years, digital imagecompletion has become a popular research topic in the fields of computer graphics and computervision, and has a very widely range of applications.At present, digital image completion techniques are mainly divided into two categories: Oneis image inpainting technology based on variation PDE; the other is image completion techniquebased on texture synthesis. This paper mainly discuss the classic algorithms of the two categoriestechnology above-mentioned: the image inpainting algorithm based on TV model and texturesynthesis image completion based on sample. On the basis of analyze and study the tworepresentative algorithms, two new algorithms are proposed in this paper. The main contents andinnovations are as follows:(1) An improved image inpainting algorithm based on TV model is proposed. In order to usemore information of known pixels closely related to the damaged pixel, this proposed algorithmuses distance of two aspects to confirm diffusion factor of neighboring pixels. One is the pixelnumber between the damaged pixel and its neighborhood pixels; the other is the distance ofdamaged pixel to the isophote of its neighboring pixels. With little pixel gap and short isophotedistance, the neighboring pixels own big diffusion factor. This algorithm can effectively minimizethe error accumulation caused due to using the known information with less relevance.(2) A new priority function is employed to calculate the priority of damaged patches.According to some deficiency in calculating the patch priority in texture synthesis imagecompletion based on sample, this paper redefines the priority function. In the new function, thestructure information of damaged patches occupies greater proportion. The data item has beenimproved, so that image completion considers the information both on isophote and gradientdirection at the same time.(3) A search strategy based on evolutionary algorithm is applied to search the optimalmatched patch. In texture synthesis image completion based on sample, the efficiency ofsearching matched patches is low and it’s easy to bring mismatch. So this paper use evolutionaryalgorithm to search the most similar matched patch in the ring area around the damaged region.And in the whole process of searching, when the gradient of the center pixel of damaged patch islarge, the size of template window is small, vice versa.

Related Dissertations

  1. Research and Application in Shop Scheduling Based on Quantum Evolutionary Algorithm,TP18
  2. Study of numerical methods for Volterra integral equations of the first kind,O241.83
  3. Research on Texture Synthesis and Transfer,TP391.41
  4. Video image segmentation based texture synthesis technology research,TP391.41
  5. Denoising problem solving variable weight fixed point algorithm,O177.91
  6. Study of Multi-objective Optimization Method Based on Evolutionary Algorithm,O221.6
  7. Research on Multimodal Function Based on Evolutionary Algorithm,TP301.6
  8. SAR Image Segmentation Based on Evolutionary Computation,TN957.52
  9. Particle swarm optimization algorithm and its application in nonlinear regression models,O212.1
  10. Research on Optimal Control for the Fermentation Process,TQ922.1
  11. Blind restoration algorithm of noisy image based on the total variation,TP391.41
  12. Implementation of Transductive Support Vector Machine in Data Prediction,TP274
  13. Research and Application on Constratined Multi-objective Evolutionary Algorithm,O224
  14. Research and Application on Evolutionary Multi-Objective Algorithm Based on Arena Principle and Niche,O221.6
  15. Image Blind Deconvolution Method Based on Gradient-Projection Algorithm,TP391.41
  16. Research on Project Logistic Alliance Management Model Based on Negotiation,F252
  17. Multiscale Total Variation Method and Its Application in Time-Lapse Seismic,O241.82
  18. Study on Immune Predictive Control Algorithm and Its Application,TP13
  19. Cardiac magnetic resonance diffusion tensor field regularization method,TP391.41
  20. Immune Mind Evolutionary Algorithm for image thresholding optimization application,TP18
  21. Research on Software-hardware Co-designing Platform Based on Evolvable Hardware,TN02

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