Edge-oriented Image Interpolation

With the development of digital image processing technology, image super-resolution reconstruction has been applied in many fields, such as the military, medical, public security, computer vision and so on, and has important prospect. Image super-resolution can be divided into two types:multiple frame super-resolution and single frame super-resolution. Super-resolution algorithm based on multiple frames not only uses the single frame image information, but also considers the similarity between different images, it is normally better than single frame super-resolution. However in many real cases, due to the various constraints, it is unable to get more frames in the same scene, the single frame super-resolution research is necessary, and gains more and more attention. This article focuses on the single frame super resolution based on the interpolation algorithm.The traditional interpolation algorithm does not consider the edge information and the image edge after processing becomes fuzzy, which affect the quality of the image after enlargement. Some interpolation algorithm consider the characteristics of the edge, but has high computation requirement and can not be arbitrarily enlarged. The vision oriented interpolation algorithm can obtain better image quality due to the particularity of image edge. An interpolation algorithm based on edge is proposed in this paper. The interpolation is computed from the surrounding pixels, which has the smoothing effect and can lead to the loss of image detail. This paper reduces the interpolation smooth effect by using the Laplace filter to enhance the image details. By calculating the dominant orientation of edge in nonflat area, the linear interpolation along the dominant direction is performed. For the flat area we direct use linear interpolation is applied, this method is simple and can be arbitrary enlarged.In order to verify the superiority of the algorithm in this paper, experiments are performed to compare the proposed method with the Gaussian process regression (GPR) algorithm, the interpolation algorithm based on iterative curvature (ICBI) and the traditional Bilinear algorithm. Image structure similarity (SSIM) and peak signal-to-noise ratio (PSNR) are selected to assess the processed image quality. Experimental results show that the proposed algorithm can maintain the edge clarity and retain the image details well; the objective evaluation criteria indicates that the proposed algorithm is superior to other algorithms.