Research On Image Restoration Based On Criminisi Algorithm And NSCT Algorithm

With the development of information technology,computer technology is becoming more and more widely used,in recent years,computer vision,digital image processing and other fields have been greatly developed,as an important technology,digital image restoration has become a hot topic for many researchers.the Criminisi algorithm,which was proposed in 2003,refers to a new height for digital image restoration,in this thesis,the principle of Criminisi algorithm is introduced,combined with nonsubsampled contourlet transform(NSCT),the Criminisi algorithm and the NSCT algorithm are analyzed and discussed,the inadequacies of the original algorithm are illustrated,and it has been improved,three correction methods are proposed for Criminisi,and domain variance method and domain energy method are introduced to NSCT algorithm.First of all,this thesis introduces the background of image restoration,research status and the development process of image restoration,explain the concept and purpose of image restoration,after that,the block repair and the criminisi algorithm are introduced in detail.Secondly,the defects of the calculation steps in the criminisi original algorithm are analyzed,in the process of determining the priority value of the point to be repaired,because of the calculation formula in the original algorithm is multiplicative,so it’s easy to have a problem with higher priority volatility,and the original algorithm is used in the calculation of the priority value,so it’s easy to have a problem with the priority fluctuation,in the priority calculation formula of the original algorithm,the confidence level(confidence)and the data item(data item)represent the pixel and the geometric relationship of the image,respectively,the two factors that affect the repair effect are due to the defect in the method of calculation,so it doesn’t work well.In view of this problem,the influence factor ? is introduced in this thesis,the product of two items in the original algorithm is changed into weighted sum,make the formula of the priority value changed from the simple multiplication to the form of self-adaptive,three kinds of function correction formulas are introduced,the original algorithm is improved.The experimental results show that the improved algorithm has achieved a significant effect.Then,this thesis introduces the nonsubsampled contourlet transform(NSCT),and an image restoration method combined with NSCT and Criminisi algorithm,the nonsubsampled contourlet transform and Pyramid decomposition are analyzed,NSCT is an image processing method,which is based on contour wave,which decomposes the image into multi layer structure and reconstructs,by analyzing the shortcomings of this method,the extraction of pixel information is insufficient,through the introduction of the neighbourhood variance,neighbourhood energy,the original algorithm is improved.The experimental results show that the improved algorithm has achieved a significant effect.Finally,this thesis makes a summary of the work and prospects for the future work.