Investigation On Image Interpolation Via Edge Information

Image interpolation is one of the basic research topics in the field of image processing.It aims to improve image resolution from a given low-resolution(LR)image to construct a high-resolution(HR)image.It has been widely applied to a variety of applications,such as image restoration,digital high-definition television application,object removal,image analysis,and image compression.Since the quality of edges has a strong impact on interpolated images,in this dissertation we mainly focus on the interpolation of image edges.Based on a comprehensive study on the existing image interpolation algorithms,several more efficient algorithms are proposed.Seriously serrated problem usually happens on image edges during the process of image interpolation.To overcome this difficulty,a new technique for rapid curve smoothing is proposed in the first part of the dissertation.By estimating the noise severity on image edges via data sampling,this technique can determine reasonable parameters to reduce noise.The proposed technique is then introduced into the process of image interpolation to smooth the jagged edges.First,edges of interpolated image are extracted.Then,the rapid curve denoising technique is employed to process the extracted edges.Finally,a HR image with high quality is obtained as the smoothed edges are embedded.This algorithm is referred to as image interpolation based on edge smoothing(ESI).In the second part of the dissertation,a new algorithm that named image interpolation with predicted gradients(PGI)is proposed.The idea of CGI is employed to diffuse the edges in the LR image firstly.Then,the unknown pixels in HR image are predicted.Finally,a 1-D directional filter is applied to processing edge pixels while a 2-D directionless filter is used to interpolate non-edge pixels.Compared to the CGI,PGI can predict the property of unknown pixels more precisely due to the strategy of gradients prediction.Compared to the usual nonlinear image interpolation algorithms,the new algorithm has a better interpretation of image edges.In the third part of the dissertation,we further develop the gradients prediction strategy and propose a new algorithm,called predictor-corrector image interpolation(PCI).It firstly uses an existing interpolation algorithm as a predictor to obtain a high-resolution image in advance.Then,the properties of each pixel in the HR image are calculated.As a result,the non-edge pixels are reserved and the edge pixels are corrected according the edge direction.Due to the predictor-corrector strategy,PCI can calculate the gradients of unknown pixels more precisely and achieve a better image interpolation result.As a whole,the ESI algorithm is a kind of explicit edge-based interpolation.It can achieve ideal result when processing binary images.However,it is not a pervasive algorithm as the quality of edges extracted from images could be unsatisfactory.The PGI and the PCI are implicit edge-based interpolation algorithms,and can be used to process every kind of images.In particular,PGI can be used in real-time interpolation and it is comparable to the existing algorithms.PCI has a high computation complexity.However,it has an even better performance with respect to most evaluation criteria of image interpolation.