Image Interpolation Based On Curvelet

With the growing demand for high-resolution image in recent years, the application of image interpolation technology is more and more widely; however, the edge effect in the image, which is interpolated by traditional method, is always unsatisfactory. Curvelet transform has a very important role in image processing and analysis due to the fact that it has a high degree of anisotropy and edge expression ability. The image interpolation algorithms, based on the nature of Curvelet and direction features of image texture, are studied in this thesis.On the basis of analyzing some typical interpolation algorithm, the basic principles of the existing image interpolation algorighm are introduced. The advantage and application prospects are analysed and the existing problems in the existing methods are summarized.In fact, the traditional interpolation methods can’t reconstruct the texture complex regional in image exactly. The Curvelet transform could express the characteristics of the edge information in image optimally. Based on Curvelet, we design a parameter named direction factor, which could estimate the direction features of image texture accurately. Experiments also confirm the relationship between direction factor and texture direction features. According to the parameters, a texture direction adaptive image interpolation algorithm for2times and3times are proposed based on Curvelet. The thought of our interpolation algorithm is:firstly some dirction factors of the image are estimated using Curvelet transform. Then, the richest feature of texture direction is selected with the biggest value of the direction factor. Finally, the interpolated points are calculated along this direction with linear weighted. We use PE (pixel error) and MPE (mean of pixel error) to compare the interpolation effect in different texture regions. Experimental results show that our method could remove the jaggy and blur at the edge effectively. Compared with the tradition methods, the interpolated image with our algorithm has a clearer edge and the superior performance in both subjective quality and visual effect.The above algorithm is proposed under the premise of a particular multiple of the image. In order to expand the scope of application, an image interpolation algorithm using neighbor weighted is proposed based on Curvelet for any real multiples. Firstly, the direction factors of the image are estimated using the Curvelet transform. As each interpolated point has four known points nearby, we compute the four weight coefficients with the direction factors and the distance between the interpolated point and the corresponding known point. The values of the interpolated point are reconstructed by means of linear weighting. Our algorithm discards the constraints that bilinear method can only be interpolated in horizontal and vertical, and it is possible to be interpolated in more direction. The algorithm proposed took into account the direction information of the image and as a result the edge direction could be located more accurately. Experimental results show that the image reconstructed with our algorithm has a similar texture trend with original image, and has a sharper edge effect as well.